119
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Discipline
An analysis of project-based work of freelancers in it sector and development of a platform to facilitate it in Turkey
Project-based work model could be advantageous for both employees and employers. In fact, there are many business benefits of project recruitment: employers can gain access to the best talented and skilled job seekers, and it can save costs and time by project-based hiring rather than full-time recruitment. In addition, employees can also benefit from project-based work by having flexible work schedules, gaining new skills and practical experience. One of the most challenging problems of fresh graduates is finding a job since they have little or no work experience. In that case, working in a project-based role could be advantageous for those inexperienced employees by helping them to get practical experience in their degree subject. This could even lead to a full-time position in the future. Nowadays, most Information Technology (IT) leaders are offering remote roles with flexible project-based opportunities. However, in Turkey, in general, project based work is difficult to arrange. IT employees face difficulties in finding and establishing contracts for project based work. The purpose of this study, as mentioned above, is to examine the current importance of project-based work and the deficiency felt in this field in Turkey, and therefore to develop a project-based job board which connects project owners with the right job seekers for their short or long term IT projects. In order to develop this project-based job platform, SQlite and DB Browser tools were used for database development purposes. The web application development was accomplished by using Django framework and Visual Studio Code as an integrated development environment.
The use of pretrained language models in sentiment analysis
Natural language processing is one of the sub-topics of linguistic science and artificial intelligence. Sentiment analysis classifies a text on any topic according to its subjective content. It is one of the methods used to examine, analyze and interpret data such as the thoughts, feelings or attitudes of individuals about a subject on various platforms. The increase in social media shares has also increased the sentiment analysis studies conducted on these platforms. Different methods are used during sentiment analysis. Classification is carried out by sentiment analysis using machine learning and natural language processing algorithms. In recent years, pre-trained language models have been used together with machine learning methods or alone. The aim of this thesis is to test the hypothetical advantages of sentiment analysis in social media comments with pre-trained language models. For this purpose, sentiment analysis was performed for Covid-19 related tweets on Twitter. Emotion intensities were determined using pre-trained language models and the results were compared. BERT, RoBERTa and BERTweet were used in the analysis. The results show that NLP techniques for sentiment analysis are as successful as other techniques.
Evaluating cognitive absorption, technology acceptance, and online educational software usability of students in online programming courses
This thesis examines the online self-learning, cognitive absorption, and technology acceptance models of students in online programming courses. This thesis discussed the role and impact of various factors on the learning of programming students. Besides investigating the effect of the educational learning-based online course content of Comp109 Introduction to Programming (Java), Comp110 Object Oriented Programming ve Comp210 Data Structures and Algorithms courses on student success, the effects of students' cognitive absorption, the acceptance, and usability of educational online learning environments were discussed. The subjects of the research are a total of 181 online students in the student group. The researcher carried out the application of the questionnaire tools to the students in the programming courses. Based on the questionnaires, this research aims to examine the effect of students' cognitive absorption, technology acceptance, and online educational software usability on student success in programming courses. In this study, the following scales were applied in the survey measurements, respectively: Cognitive absorption scale (13 items), technology acceptance model (14 items), online educational software usability (39 items). The preparation, ethical approval and implementation of the relevant survey and data analysis research took three months. This study evaluates students' cognitive assimilation, technology acceptance, and e-education software usability scale and suitability for online education in online programming courses. Each of these three scales was examined from seven different perspectives: Gender, Course Code, Department, Age, Expertise, Frequency of Internet use, and Interest in technology. Normality test, ANOVA, Mann-Whitney U Test and Kruskal Wallis Test were used to compare and evaluate these measurements. In addition to these test measurements, the MANOVA test was also applied for multivariate analysis. MANOVA test was carried out in detail from 4 different perspectives. The results of each of these scales were investigated from seven different perspectives and compared with related studies, and suggestions were presented based on the obtained results. In line with my findings, the positive effect of online programming courses on cognitive absorption has been demonstrated. However, it was also shown that there was no significant difference between demographic variables. In addition to these, a literature search was made for all three scales for discussion and a discussion was written. Results are written according to the discussion and analysis.
Customer transaction predictive modeling via machine learning algorithms
Bu çalışmanın temel amacı faktoring sektöründe aktif olarak faaliyet gösteren bir şirketin müşterilerinin davranışlarını ve özelliklerini saptayabilmek, buna bağlı olarak da müşterilerin geçmiş verilerinden yola çıkarak, keşifçi veri analiziyle ölçülebilir parametreler yakalayabilmek ve akabinde hedefe yönelik tahminsel modellemeler gerçekleştirebilmektir. Sınıflandırma modeli algoritmalarından SVM ve Extra Trees modellerinde %80 seviyesi üzerinde isabet oranı yakalanmıştır. Bu sayede yüksek potansiyel gösteren, yani sorgulattığı çeki, işletme olarak onayladıktan ve teklif yapıldıktan sonra kabul etme ihtimali yüksek olan müşterileri tahmin edilmesi daha etkin, verimli ve doğru yaklaşımlar içerisinde hareket edip aksiyon alarak, işletme bazında işlem hacmine doğrudan katkısı sağlanması amaçlanmıştır.
Fraud detection and prediction with machine learning applications
The main purpose of this study is to determine the fraudulent activities on transactions of the customers of a company that is active in the factoring sector, and accordingly, to capture measurable parameters with exploratory data analysis based on the historical transaction and connection data of the customers, and then to perform predictive models for the target. A hit rate of around 79% was achieved in XGBoost and CATBoost models, which are classification model algorithms. In this way, it is aimed to directly detect fraudulent activities on a trasnaction basis by acting in a more effective, efficient and correct approach after detecting the customer that shows high potential to make fraud.
Customer churn prediction for the Pay-TV sector
Understanding the reasons for customer churn provides added value in terms of retaining existing customers, as customer attrition leads to revenue loss for companies and incurs marketing costs for acquiring new customers. In this study, the 6-month historical data of a Pay-TV company operating in Turkey was used, and due to the imbalanced nature of the dataset on a label basis, the oversampling method was applied. During the model development phase, various artificial learning algorithms (Random Forest, Logistic Regression, K-Nearest Neighbors, Decision Tree, AdaBoost, XGBoost, Extra Tree Classifier) were utilized, and their performances were compared. Based on the evaluation of success criteria for each model, it was observed that the tree-based Random Forest, Extra Tree Classifier and XGBoost achieved the highest performance for this dataset.
A web of things system to control party equipment
This thesis focuses on the equipment needed to throw a successful party and how to coordinate and synchronize these components seamlessly. It also explores the integration of the Web of Things (WoT) paradigm into party organization to improve coordination and communication between entertainment devices. The research aims to provide a detailed description of the necessary equipment and presents innovative applications of the WoT paradigm to enhance the coordination and synchronization of entertainment elements. By doing so, this thesis contributes to the evolving field of WoT and its potential applications in dynamic and interactive environments such as parties.
Comparing the effectiveness of graph neural networks and machine learning algorithms for fNIRS-based neuromarketing research
Functional near-infrared spectroscopy (fNIRS) has some advantages over other brain imaging methods in terms of cost and portability. For this reason, its use in neuromarketing is increasing. However, fNIRS brings some challenges along with its advantages. Due to features such as multichannel measurement and high temporal resolution, the nature of fNIRS data is complex and multidimensional [7]. Neuromarketing researchers have utilized machine learning algorithms to overcome these challenges. When these studies are analyzed, it is seen that successful results have emerged. Machine learning has influenced researchers working on graphs as well as neuromarketing researchers. Thus, graph neural networks have emerged, which allow the application of artificial neural networks to graph data structures. Thanks to the fact that the brain can be modeled as a graph structure using functional connections [14] and the high temporal resolution of fNIRS [7], there are neuroimaging studies using graph neural networks and fNIRS together. However, despite successful results, there is no neuromarketing research using this combination. Therefore, in this study, the performance of graph neural networks in fNIRS-based neuromarketing was analyzed and compared with machine learning algorithms that have been shown to yield successful results in this context. For the comparison, fNIRS measurements of a neuromarketing experiment conducted to determine perceptions toward brands were used. In the experiment, consumers were asked to decide (yes/no) whether the adjective shown with the brand logo was appropriate for the brand. The data set was obtained by cleaning the obtained measurements. First, a supervised machine learning approach was applied to this dataset. After the dataset went through several data preprocessing stages, various algorithms were trained on it. These were K-Nearest Neighbors, Support Vector Machines, Random Forest, Naive Bayes, and XGBoost algorithms. Then, two different voting classifiers, one for soft voting and one for hard voting, were created from the algorithms that were more successful than the others. After the machine learning approach was completed, the graph neural network approach was applied. The data obtained through fNIRS was transformed into a graph structure using functional connections in the brain. The Pearson correlation coefficient was used to calculate the functional connections. Since a graph was created for each trial of the participants and each graph had a label (yes/no), classification was performed at the graph level. For graph classification, the generated graphs were given as input to graph neural network architectures. The architectures used in the study consisted of Graph Convolutional Network, Graph Attention Network, and Graph Isomorphism Network. Finally, a soft voting classifier was created by combining these architectures. Test accuracy values of all methods were calculated and binomial confidence intervals were added to these values. The comparison results showed that machine learning algorithms generally outperform graph neural networks. Additionally, machine learning models based on ensemble learning have the best scores.
Customer churn prediction for a personal care product retail chain operating in Turkey
Understanding the reasons for customer loss and the customer behaviors leading to it, as well as being able to predict customer's loyalty to an industry or a company provides enormous advantages in retaining existing customers and avoiding revenue loss due to the marketing and advertising costs associated with attracting new customers. In this study, the 29-month data from a personal care product retail chain operating in Turkey was used, and because of the imbalanced values and non-customer entries of the dataset, the oversampling method and synthetic sampling was applied. During the model development phase, Logistic Regression, Decision Tree, K-Nearest Neighbors, Random Forest, Extra Trees Classifier, and MLP (Multi-Layer Perceptron) Classifier were applied, and their performances were evaluated using metrics such as accuracy, recall, F1-score, precision, and confusion matrix. Based on these comparisons, it was observed that the Random Forest and MLP Classifier models demonstrated the best performances for this dataset, while other tree-based algorithms, such as the Extra Trees Classifier and Decision Tree, achieved slightly lower but comparable performance.
Customer-merchant shopping behavior modeling in e-commerce
The rapid advancement of e-commerce has necessitated the adoption of sophisticated analytical models to decipher customer behaviors and evaluate seller performance. In this study, we propose a novel 'Customer-Seller Shopping Behavior Model' that integrates customer preferences, purchasing patterns, and seller product offerings. The model utilizes customer metrics such as RFM (Recency, Frequency, Monetary), category preferences, and customer lifetime value, alongside seller data on category focus and sales performance, to facilitate segmentation and personalized recommendations. Employing K-Means, hierarchical clustering, and decision tree algorithms, customers and sellers are categorized based on their behavioral similarities. Furthermore, predictive models and hybrid recommendation systems are implemented to forecast customer purchasing propensities and match them with suitable sellers. The results demonstrate that the proposed model significantly enhances customer engagement, seller effectiveness, and overall platform efficiency. This research contributes to a data-driven, scalable, and dynamic analytical framework for e-commerce platforms.
A two-tower retrieval approach for personalized product recommendations in cosmetic e-commerce
Recommender systems play a vital role in enhancing user experience and driving engagement in e-commerce platforms. This thesis presents a comparative analysis of two personalized product recommendation approaches using retrieval- based neural architectures implemented with TensorFlow Recommenders (TFRS). The first approach relies solely on ID-based embeddings using user and product identifiers. The second model enriches product representations by incorporating semantic content features, specifically product title and upper category information. Both models are trained using implicit purchase data and evaluated with the FactorizedTopK metric, which estimates recall@K through nearest-neighbor retrieval. The study assumes a warm-start scenario and emphasizes the value of semantic context in embedding space for product retrieval. Our findings highlight the benefit of including content features for improving recommendation accuracy, particularly in contexts where semantic similarity between products is relevant.
Deep learning for predicting soil properties and climate change
Soil is a fundamental natural resource that not only provides essential nutrients, minerals, and elements necessary for plant growth but also plays a vital role in climate regulation and broader ecosystem functioning. This master's thesis aims to improve the practises of modern Artificial Intelligence (AI) methodologies in Digital Soil Mapping (DSM) practices across Türkiye by integrating key environmental variables such as land cover, topography, climate data, and other spatial factors. To achieve this objective, a comparative analysis of seven different machine learning models was conducted to evaluate their performance in predicting soil properties based on these variables.
Anayasa Mahkemesi Kararlarının Simülasyonu: Türk Bireysel Başvuruları İçin Çok Etmenli Bir Büyük Dil Modeli (LLM) Çerçevesi
This research examines the ability of large language models (LLMs) to emulate judicial decision-making in constitutional court cases. We used three cutting-edge models—GPT-5, Gemini, and Claude—to look at 343 decisions made by the Turkish Constitutional Court between 2014 and 2024. We did this using a two-stage evaluation framework that mirrored how courts really work. The study evaluated model efficacy in both admissibility determinations and substantive rights infringement judgments. During the admissibility stage (Stage 1), the accuracy rates varied from 68.80% for Claude to 81.34% for GPT-5, with majority voting achieving 79.59%. GPT-5 had the fewest total mistakes and a balanced approach, while Gemini and Claude were more likely to think that something was not admissible. On the other hand, courts had a more moderate acceptance rate. In Stage 2, when rights were violated, all three models had the same accuracy of 81.50%. However, majority voting did better, with an accuracy of 83.24%. In this case, GPT-5 tended to have partial matches, Gemini had the most exact matches, and Claude was in the middle, showing that each had different strengths in legal reasoning. Patterns of inter-model agreement showed that there was a lot of convergence, but it wasn't always the same. In Stage 1, unanimous agreements, though less common, had the highest accuracy (87.32%), showing that consensus decisions are reliable. In Stage 2, GPT-5 and Claude were the most in line with each other (88.52%). These results indicate that ensemble methods and hybrid human–AI approaches could improve the consistency and robustness of judicial decision-making. The results show that even general-purpose LLMs can understand complicated constitutional principles and come up with structured, court-like reasoning that is serious enough to be used in legal situations. Although existing models struggle to replicate the comprehensive intricacies of judicial reasoning, the persistent superiority of ensemble methodologies suggests that specialized legal AI systems may exceed general models, potentially revolutionizing constitutional jurisprudence by improving efficiency, consistency, and accessibility to justice
Image classification of customs procedure documents using machine learning and deep learning models
The images of documents used in customs procedures were classified using machine learning and deep learning models. As image classification methods, Gaussian Naive Bayes, Random Forest, and Support Vector Machine were used in the machine learning field, and Convolutional Neural Network, Vision Transformer, ConvNext and EfficientNetV2 were used in the deep learning field. Datasets were organized specifically for each model type, and models were created with parameter values determined to ensure the models worked at the optimum level. Training and testing were conducted with the relevant datasets. The models were compared in terms of training times and accuracy rates. The best results were achieved with Convolutional Neural Network, which had a training time of approximately 11 minutes and an accuracy rate of 98,32%; second-best accuracy result achieved with Random Forest, which had an accuracy rate of 97,94% and second-best overall results achieved with EfficientNetV2, which had a training time of approximately 22 minutes and an accuracy rate of 95,89%.
Enhancing intrusion detection with privacy-preserving federated learning: Differential privacy and incremental learning integrating
In cybersecurity, Intrusion detection systems (IDS) are crucial in scanning networks and systems to detect malicious activities and help to identify the threats before sensitive data is compromised. The introduction of machine learning (ML) has enhanced IDS by providing mechanisms for automated and intelligent threat detection. However, the training of ML models in such distributed settings, e.g., in Federated Learning (FL), may be an issue for training and may still expose sensitive information through model parameters analysis. FL mitigates certain privacy issues by localizing the data, however, it is not sufficient for genuine privacy protection and requires improvement. Specifically, Incremental Learning (IL) improves IDS providing the ability models adjusting to new cybersecurity threats without the need for retraining from scratch. This keeps it computationally inexpensive and allows it to quickly adapt to novel attack behaviours. In particular, we propose Federated Differential Privacy Enhanced Model Aggregation, a method targeted towards enhancing both privacy and accuracy in federated ML context. It uses a server-client architecture, where a global model is initialized and further improved with client-side training, while updates are securely aggregated. We also used a multi-layer perceptron trained with the DP-SGD optimizer which drops noise into the gradients to achieve better data privacy. We evaluated the performance, the experimental results show that the accuracy of class incremental learning in our proposed approach can reach 92.4% and feature incremental learning can reach 99.4%, they demonstrate that the learning can match the new data well. The procedure remains privacy-preserving and efficient and has good performance over diverse dataset types, so we believe it to be a valid candidate for intrusion detection systems (IDS) in modernity.
Klasik soru değerlendirme karar destek sistemi
Bu tez çalışmasında, öğreticilerin klasik soru değerlendirme süreçlerinde kullanmaları için not önerisinde bulunacak bir karar destek sistemi geliştirilmiştir. Bu çalışmanın amacı öğreticilerin kendi not değerlendirmeleri ile sistemin not değerlendirmeleri arasında karşılaştırma yaparak adil bir değerlendirme süreci ortaya çıkarmalarına yardımcı olmaktır. Bu çalışmada klasik soru cevaplarını oluşturan cümlelerin kelimelere ve bu kelimelerin köklerine ayrılması için, aynı zamanda kelime yazımlarının denetlenmesi ve olası düzeltmeler için Zemberek DDİ Kütüphanesi kullanılmıştır. Her bir klasik sorunun doğru cevabı ve öğrenci tarafından verilen cevabın benzerliğini hesaplamak ve buna göre önerilecek notun hesaplamasını yapmak için ise Vektör Uzay modeli ve Kosinüs Benzerliği Teoremi kullanılmıştır. Geliştirilen algoritma sonuçlarının bilimsel geçerliliği, öğreticinin not değerlendirmesi ve sistemin not değerlendirmesi arasındaki ilgileşim (korelasyon) hesaplanarak ispatlanmıştır. Korelasyon hesaplamaları için Pearson's Korelasyon katsayısı (Pearson's Correlation) ve T-test yöntemleri kullanılmıştır. Yapılan analizler sonucunda Pearson's Korelasyon katsayısı 0,667 olarak bulunmuştur. T-test analizi sonucunda da Pearson's Korelasyon katsayısı hesaplamaları ile birebir benzer sonuçlar elde edilmiştir. Bu sonuçlar sistemin not değerlendirmeleri ve öğreticinin not değerlendirmeleri arasında güçlü bir korelasyon olduğunu göstermiştir.
Dikkat seviyesinin gerçek zamanda izlenmesi; bir web uygulaması
Dikkat seviyesinin ölçüldüğü ve izlendiği bu tez kapsamında; EEG ölçümleri Emotiv Epoc Headset ile yapılmış ve veriler gerçek zamanlı olarak veritabanına kaydedilmiştir. Kaydedilen frekans verileri FFT (Fast Fourier Transform) yöntemi ile filtrelenmiş ve bu filtreleme yöntemi bizi, elde etmek istediğimiz beta, teta, alfa, delta v.b. frekans aralıklarına ulaştırmıştır. Çalıştırdığımız bu filtre ayrıca, kişinin kullandığı bilgisayar üzerinde yüklü olan küçük bir uygulamanın içinde mevcuttur. Sahip olduğu bu filtre, kişinin duygu tipi ve stres düzeyini belirleyebilen aynı zamanda bilimsel normlara göre veri analizi yaparak kişinin, dikkat seviyesi gerçek zamanlı olarak ölçülebilir hale getirilmiştir. Ayrıca ölçülen frekansların ve alınan verilerin web arayüzünün yönetim paneli üzerinden, zamana bazlı frekans bilgilerine ulaşabildiğimiz ve grafik olarak da gösterebildiğimiz yönetici sayfası bulunmaktadır. Bu sayfadan her kullanıcının stres düzeyi ve duygu tipini öğrenebildiğimiz gibi, acil bir durumda kişiye hızlı bir şekilde müdahale etmemizi sağlayabilmektedir.
Optimizing network performance rough image processing techniques in omputer science it
A fundamental apparatus for exploring digital protection risks in Internet of Things (IoT) networks is the Botnet of Things (BoT-IoT) dataset. This examination proposes an original way to deal with order BoT-IoT information by changing over the crude information into RGB images and afterward applying calculations, for example, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Irregular Woodland to a helpful classification model. The BoT-IoT dataset is at first changed over into RGB images, where each component vector addresses a pixel in the image. To protect the spatial data of the first information, these photographs are in this manner saved in the.png design. We might take benefit of the strong capacities of example acknowledgment and image regulation procedures for IoT information examination with this change interaction. The RGB images delivered from the BoT-IoT dataset were then sorted utilizing an aggregate classification method. The ensemble includes the most striking Irregular Woodland classifiers, KNN, and SVM, every one of which contributes unmistakable qualities to the general classification issue. To show up at the last classification choice, the estimates of discrete classifiers are shared utilizing a weighted democratic system. The recommended strategy is compelling, as proven by trial discoveries on the BoT-IoT dataset, where the ensemble model accomplishes a great exactness of generally 99.36%. What's more, the model performs well with regards to precision, review, and F1-score, exhibiting its adequacy in isolating noxious action from harmless movement in Internet of Things networks. Inside and out review was likewise finished to explain every classifier's commitment inside the ensemble setting. This examination uncovers the fitting properties of the classifiers: Arbitrary Timberland consolidates the limit of ensemble learning for upgraded speculation, KNN makes utilization of nearby neighbourhood data, and SVM succeeds at dealing with testing choice limits. In any case, this work offers a new way to deal with BoT-IoT information recording using ensemble learning strategies and RGB image models. The extraordinary exactness achieved features this strategy's capability to improve digital protection in Internet of Things networks. Through effective distinguishing proof and alleviation of pernicious activities, the proposed approach adds to the advancement of IoT security exploration and safeguards networked frameworks from arising digital threats.
Discovering sustainable ecosystem with green iot
Internet of things (IoT) is one of the most extensively used concepts that enables large numbers of connected devices to obtain data, analyse it, and utilize it. However, the increased connectivity of objects and devices in IoT brings serious environmental problems, namely those connected with power consumption and e-waste generation. This paper aims to examine the possibility of "Green IoT" which encompasses using efficient energy technologies, recycled/reusable materials, and green energy to reduce the adverse effects IoT could cause to the environment. Case studies and real-world applications are used in the research where energy-efficient designs of Green IoT technologies are studied to show that Green IoT technologies can drastically minimize energy use, control emissions, and promote a more sustainable IoT environment. Several papers have been studied and some of the areas of contribution include enabling technologies which include Green Radio Frequency Identification (RFID), energy harvester devices, low power communication protocol, all of which enable improvement of energy efficiency in the IoT systems across smart cities, healthcare, and industrial automation. Furthermore, this paper outlines key issues arising from the application of Green IoT, including lack of protocol coordination and ineffective policy measures. The results confirm Green IoT as a viable solution regarding concerns over the lack of sustainability in IoT interventions and provide potential directions and strategies for policy makers and executives on how to create a less damaging footprint of IoT while making the most of the opportunities such technologies offer for the improvement of the quality of life and the creation of sustainable development.
Meslek liselerinde okul aile işbirliğini arttırmada bilgi teknolojisi kullanmanın etkisi: Karşılaştırmalı bir araştırma
Bu araştırmada amaç, meslek liselerinde okul aile işbirliğini güçlendirmede bilgi teknolojisi kullanımının katkısını incelemektir. Araştırmada elde edilen veriler SPSS 21 paket programında, frekans, yüzde analizi, ön test ve son test teknikleri ile analiz edilmiştir. Araştırma sonucunda; okul aile işbirliğini güçlendirmede bilgi teknolojisi kullanımının öğretmen, veli, okul yönetimi ve öğrenci arasında pozitif yönde bir katkısı olduğu; özellikle velilerdeki okul yönetime karşı olan önyargıları ortadan kaldırdığı ortaya çıkmıştır.
Mobil cihazların (tablet pc)eğitim/öğretime etkisinin belirlenmesi
Bu araştırmada Tablet PC'lerin Eğitim/Öğretime etkisinin belirlenmesi amaçlanmaktadır. Bu amaç çerçevesinde oluşturulan ölçekler yardımıyla veri toplama işlemleri gerçekleştirilmiştir. Tablet PC'lerin eğitim öğretime etkisinin araştırılması ile günümüzde hızla yaygınlaşan mobil cihazların eğitim sistemine dahil edilmesi yoluyla, bilimsel ve toplumsal açıdan öğrencilerin okula, eğitim ve öğretime olan ilgisinin artacağı düşünülmüştür. Bu araştırmanın mobil cihazların eğitimde kullanılması bağlamında öğretmen-öğrenci, öğrenci-öğretmen, öğrenci-öğrenci, öğretmen-öğretmen arasında etkileşim ve dayanışma kültürünün oluşturulması açısından yarar sağlaması amaçlanmaktadır. Araştırmada dört adet ölçek, Amasya il merkezinde Tablet PC dağıtımı yapılmış olan Amasya Anadolu Lisesi ve Amasya Anadolu İmam Hatip Lisesi'nde öğrenim gören 10'uncu sınıf öğrencilerine uygulanmıştır. Tutum ölçeği, "Tablet PC'nin Derslerde Kullanımına Yönelik Öğrenci Tutumlarını" ölçmek amacıyla, Tablet PC kullanan 216 öğrenciye uygulanmıştır. Tutum ölçekleri sonucuna göre araştırmaya katılan lise öğrencileri genel olarak Tablet PC'nin derslerde kullanılıp kullanılmaması konusunda kararsız görüş belirtmişlerdir. Gözlem formları ve başarı testleri, deney grubu (Tablet PC'li) ve kontrol grubu (Tablet PC'siz) olarak ayrıştırılan sınıflara ayrı ayrı uygulanarak iki grup arasındaki fark bulunmaya çalışılmıştır. Gözlem formu, deney ve kontrol gruplarında sözel derslerden seçilen "Dil Anlatım" ve sayısal derslerden seçilen "Matematik" derslerinde belirtilen konu ve üniteler ders öğretmeni tarafından anlatılırken araştırmacı tarafından doldurulmuştur. Gözlem formu sonuçlarına göre, deney grubundaki (Tablet PC'li) ders anlatım sürelerinin, kontrol grubundaki (Tablet PC'siz) ders anlatım sürelerinden kısa olduğu belirlenmiştir. Başarı testleri deney ve kontrol gruplarına, "Dil Anlatım" ve "Matematik" derslerinde konu anlatımları yapılmadan önce Ön-Test olarak uygulanmış, konu anlatımları yapıldıktan sonra Son-Test olarak uygulanmıştır. Başarı testleri sonuçlarına göre deney grubundaki (Tablet PC'li) Ön-Test ve Son-Testler arasındaki farklar, kontrol grubundaki (Tablet PC'siz) Ön-Test ve Son-Testler arasındaki farklardan önemsenmeyecek derecede fazladır.
Siber suçlar üzerine bir araştırma afyonkarahisar örneği
Bilgi iletişim teknolojilerinin ve bu teknolojilere erişilebilirliğin her geçen gün artmasına paralel olarak bu sistemlere yönelik suç türleri de artmaktadır. Sanal bir ortamda paylaşılan bilgiler bilinçsiz veya yetersiz güvenlik adımları uygulamalarıyla kötü niyetli kullanıcıların eline geçebilmektedir. İnternet kullanıcıları bazı mağduriyetler yaşayarak bilmeden adli bir konunun tarafı olabilmektedir. Bu noktada siber suçun insanlar tarafından nasıl algılandığı son derece önemlidir. Bu çalışmada Afyonkarahisar ölçeğinde lise öğrencilerinden yükseköğretim öğrencilerine ve kamu çalışanlarına kadar siber suç algısının temel düzeyde ölçümü yapılmıştır. Afyonkarahisar örnekleminde kamu çalışanları ile öğrenim gören öğrencilere yönelik internet kullanım alışkanlıklarının tespiti, siber (bilişim) suça ilişkin görüşlerinin tesbiti ve suç farkındalıklarının ortaya konması amacıyla bir anket çalışması yapılmıştır. Anket sonuçları yorumlanarak katılımcıların internet kullanım araçları, amaçları, alışkanlıkları, siber suç bilgileri, tehdit algıları ve farkındalıkları, internet ortamında bireysel özgürlüklerin kullanımı ve korunması, güvenlik uygulamalarına yaklaşımları ve duyarlılıkları hakkında bir düzey ölçümü yapılmıştır.
Phishing website detection based on a novel artificial intelligence technique
Due to the increasing use of Internet, researchers widely study to prevent cybercrimes. One of the cybercrimes is phishing websites. As cybercriminals employ increasingly sophisticated tactics, there is a growing need for innovative approaches to detect and prevent these threats. In this thesis, the time-varying mirrored S-shaped transfer function is applied for BGWO, BPSO, and BHHO, and the proposed models are performed on the phishing websites dataset. The proposed models have promising results on the phishing websites dataset
Şarkı sözlerinden yeni şarkı türetme: Derin öğrenme yöntemleri ile sanatsal üretim
Şarkı sözü yazımı yaratıcı bir süreçtir. Geleneksel olarak bu süreç ilham, sezgi ve bireysel yaratıcılığa dayanır. Ancak yapay zekâ teknolojilerinin gelişmesiyle birlikte bu süreç yeni bir döneme girmiştir. Son yıllarda, özellikle yaratıcı metin üretimi alanında, Doğal Dil İşleme (NLP) kapsamında önemli ilerlemeler kaydedilmiştir. Özellikle GAN mimarileri, özgünlük ve rastlantısallık gerektiren içeriklerin üretiminde dikkat çekmektedir. Bununla birlikte, GAN'ların metin üretimine uyarlanması, görüntü üretimine kıyasla daha karmaşık bir süreçtir. Türkçe gibi sondan eklemeli dillerde hem anlamlı hem de yapısal olarak tutarlı metinler üretmek daha da zordur. TextGAN, SeqGAN, LeakGAN, RankGAN, MaliGAN ve RelGAN gibi mimariler, Türkçe şarkı sözü üretimi bağlamında sistematik olarak değerlendirilmemiştir. Bu durum, çalışmanın temel problemini oluşturmaktadır. Bu tez çalışmasının amacı, mevcut şarkı sözlerini analiz ederek, özgün ve sanatsal değeri yüksek içerikler üretebilen en uygun yapay zekâ modelini belirlemektir. Bu kapsamda, derin öğrenme modellerinin yanı sıra duygu analizi, ritmik yapı analizi ve yaratıcı metin üretimi teknikleri kullanılmıştır. Elde edilen sonuçlara göre, düşük BLEU ve Jaccard skorları, modellerin referans metinlerle yüzeysel düzeyde benzerlik kurmakta zorlandığını göstermektedir. Buna karşılık, RelGAN modeli Cosine Similarity metriğinde en yüksek skoru (0.0693) elde ederek, anlamsal olarak en tutarlı çıktıları üretmiştir. MaliGAN modeli ise düşük perplexity değeri (9588.23) ile dil modelleme açısından öne çıkmıştır. Ancak, çeşitlilik metriklerinin (TTR ve Unique Word Ratio) 1.0 değerine ulaşması, anlamlı tekrarların eksikliğini ve doğal akışın zayıf olduğunu göstermektedir. Edebi analizlere göre, yalnızca RelGAN modeli tematik bütünlük ve duygu yoğunluğu açısından tatmin edici şarkı sözleri üretebilmiştir. Sonuç olarak sayısal başarı ve edebi değerlendirme birlikte ele alındığında, RelGAN en dengeli ve başarılı model olarak öne çıkmıştır. Türkçe gibi eklemeli yapıya sahip dillerde şarkı üretimi için GAN modelleri hâlen sınırlı kalmakta olup, gelecekte Transformer tabanlı yapılarla desteklenmesi önerilmektedir.
Renkli görüntüler için kriptografik şema
Due to the extensive use of multimedia applications, image coding standards are one of the ways that supported multimedia spreading. Internet is full of digital documents that are uploaded by thousands every second through social media or any websites for commercials or business. Encryption and decryption techniques are required for digital multimedia distribution. Today, privacy and authentication services are crucial for data dissemination through the Internet. This work focuses on image security integrating encryption with multimedia compression systems. Also, image encryption algorithms combining with JPEG encoding are also proposed. Three approaches for integrating encryption with image compression system are proposed. These are Color Plane Permutation, Discrete Cosine Transform (DCT) coefficients, and sign encryption of DCT coefficients. The proposed approaches are supported by selective encryption, are a portion of the coefficients from either the final results or intermediate steps of a compression system. The securities of the proposed approaches are also analyzed. Simulations results show that the proposed system is secure, low cost, and compatible with direct bit-rate control. These properties make the study suitable for multimedia applications with the real-time operation and image transmission.
Kablosuz sensör ağlarında saldırı tespiti
Given the possibility of using intelligent agents for the development of intrusion detection systems, obtained through a bibliographic survey, this work will first seek to develop the following hypothesis: The use of intelligent agents in the development of intrusion detection and prevention solutions in wireless sensor networks is quite suitable for such equipment, as they favor the development of solutions. In addition, the use of intelligent agents allows a better development of these systems, to provide satisfactory computational performance, adaptability, scalability, shorter time latency and load balancing. The general objective of this paper is to propose a framework, based on agents, for detecting intrusion in wireless sensor networks, with the best possible adaptation resources of network devices
Bireyleri, kuruluşları ve toplumun davranış kalıplarını verilerin analiz edilmesi ve jeosuzayal bilgilere dayanarak çözümler bulması için güçlendirecek büyük veri odaklı elektronik devlet
In the information era, it is more crucial than ever to use big data effectively in order to increase the efficiency, accountability, and openness of electronic government (e-Government). This essay investigates the revolutionary potential of big data-driven e-Government in altering societal behavioral patterns in Turkey. We aim to provide examples of how e-Government may employ these technologies to improve public services, inform decision-making, and actively engage citizens in governmental activities. Modern big data analytics, machine learning techniques, and predictive models are used to achieve this. This research examines the various aspects of big data applications in Turkish e-Government, such as how they support tailored public services, efficient resource management, sound policy-making, and effective citizen participation. We go into more detail on how these applications affect people's and organizations' behavioral habits. The findings indicate that big data-driven e-Government holds great promise for rethinking the citizen-government interaction model and fostering a society that is better data-informed in Turkey. As a result, this might promote organizational performance, foster the development of new social norms in communities, and make it possible for individuals to participate actively and meaningfully in governance in Turkey. The study emphasizes how careful management and ethical use of big data in e-Government can contribute to building a more inclusive, transparent, and ABSTRACT BIG DATA-DRIVEN ELECTRONIC GOVERNMENT TO EMPOWER INDIVIDUALS, ORGANIZATIONS AND SOCIETY BEHAVIORAL PATTERNS TO ANALYZING DATA AND FINDING SOLUTIONS BASED ON THE GEOSPATIAL INFORMATION ALI, Ali Hussein Ali M.S., Electrical and Computer Engineering, Altınbaş University, Supervisor: Asst. Prof. Dr. Sefer Kurnaz Date: June / 2024 Pages: 54 viii participatory government while being aware of the challenges that are inevitably there in Turkey.
Akıllı araç görüşünde derin evrişimli sinir ağı kullanılarak şerit segmentasyonu ve yol tespiti
This thesis will delve into the intricacies of designing, training, and evaluating a deep CNN architecture for lane segmentation and subsequently harness the discriminative power of SVMs for road detection. Through a comprehensive investigation of these methodologies, this thesis aims to contribute to the advancement of smart car vision systems, ultimately paving the way for safer and more reliable autonomous driving solutions in an ever-evolving urban landscape. As society becomes increasingly reliant on digital communication and data, the ability to understand, process, and generate human language has become a critical component in various applications, from virtual assistants and sentiment analysis to machine translation and content recommendation systems.
Değiştirilmiş KNN algoritmasını ve özellik ağırlıklandırmayı kullanarak DNS trafiği tabanlı izinsiz giriş tespiti
All computer networks can be the target of attacks, from the smallest, such as home networks, to the largest and most complex, such as corporate networks. To make the situation even worse, programs can easily be obtained through the Internet (or by other means), and then used for malicious actions, making any inexperienced person have the same skills as a professional It is necessary to propose a new database that contains a portion of the traffic that is present on an existing actual network environment in order to carry out testing in a real environment. This is necessary in order to accomplish the goal of conducting testing in a real environment The main objective is to create a network user control mechanism in order to recognize the user's legitimacy through their actions.
Öğrenci devamının kaydedilmesi ve tanınması derin öğrenmeyi kullanan yüzleri
There are several methods available to monitor student attendance in classes, such as biometric, radiofrequency, face recognition, and paper-based systems. However, the face recognition-based approach has been found to be both efficient and secure. In this study, a threshold to confidence has been implemented through Euclidean distance values to enhance the identification process. The Local Binary Pattern Histogram (LBPH) algorithm has been utilized for this purpose, as it has been demonstrated to be more effective than other methods such as Eigenfaces and Fisher faces. The Haar cascade method has been used for facial detection due to its robustness. The system's performance has been assessed in various scenarios, including recognition rates, false-positive rates, and detecting unknown individuals with or without a threshold. The system has demonstrated an impressive 79% recognition rate for students, with a 24% false-positive rate, and can identify students wearing glasses or a beard. The LBPH algorithm and Haar Cascade method contribute to the system's exceptional performance. The recognition rate for unregistered individuals in facial recognition technology is noteworthy even without the use of a threshold value, sitting at a commendable 64%. Moreover, the rate of false positives is impressively low, remaining at approximately 15% and 31%.
Elektrokardiyogram kayıtlarına dayanarak kalp hastalıklarının sınıflandırılmasına yönelik sinir ağları
The precise and timely detection of cardiac arrhythmias is crucial for healthcare practitioners since it significantly impacts patient outcomes. This study is centered on the enhancement of electrocardiogram (ECG) signal classification, with a specific emphasis on deep learning and the proposed model. The complex patterns included in electrocardiogram (ECG) data, commonly used in clinical practice, pose challenges to conventional classification methods. The efficacy of the technique is evidenced by its ability to generate notable results. The classifier has been trained and validated using a comprehensive electrocardiogram (ECG) dataset that has undergone preprocessing. The performance measures (accuracy, precision, recall) underscore the exceptional ability of the system to identify cardiac arrhythmias. The validation of the model on an independent dataset reveals its capacity to generalize, maintaining a high level of accuracy and providing valuable insights into arrhythmia. The findings are supported by a comprehensive classification report, confusion matrices, and ROC analysis. This work showcases the potential of deep learning in transforming the detection of cardiac arrhythmias, using the model as an illustrative example. The results of this study contribute to the enhancement of ECG classification techniques, hence enhancing the accuracy and reliability of diagnoses and patient treatment in the field of cardiology.
Optik fiberlerin ve modernin katkıları internet üzerinde öğretimi geliştirme teknikleri corona Covid-19 pandemisi sırasında
Internet service has become an indispensable aspect of daily life, extending its necessity to homes, institutions, and organizations worldwide. With the surge in technological advancements, increased demand, and a growing number of internet users, traditional copper wires have proven inadequate to meet the evolving needs. The progression of internet technology from the first to the fifth generation underscores the imperative to adapt and keep pace with developed nations. In response to these demands, optical fibers have emerged as a pivotal solution, providing high-speed data transmission essential for millions of users globally. This is particularly crucial given the substantial data volumes, often exceeding hundreds of gigabytes. Optical fibers find diverse applications, with Fiber To The Home (FTTH) standing out as a specialized conduit for efficiently transferring internet services from companies to homes and workplaces. Moreover, optical fibers play a pivotal role in the deployment of submarine and terrestrial cables, fostering global connectivity through the World Wide Web (WWW). This not only ensures seamless data flow but also supports critical applications such as Fiber To The Home, connecting nations and facilitating robust internet services worldwide. The shift towards optical fibers is imperative in the context of evolving internet technology, with the fifth generation becoming increasingly prevalent. This transition demands the adoption of means, media, and high-speed transmission lines capable of handling the demands of modern internet usage. Optical fibers, with their capacity for high-speed data flow, emerge as a key enabler in this scenario. In conclusion, the utilization of optical fibers is instrumental in addressing the escalating demands for high- vii speed and reliable internet services. From FTTH applications to the establishment of global connectivity through submarine and terrestrial cables, optical fibers play a central role in meeting the evolving needs of the digital age.
Kardiyovasküler hastalık tahmininin geliştirilmesi için entegratif makine öğrenmesi yaklaşımları: XGBoost ve ANFIS algoritmalarının karşılaştırmalı analizi
Cardiovascular diseases (CVDs) are the leading cause of death globally, underscoring the need for advanced detection and diagnostic methods to enhance patient outcomes. This study investigates the efficacy of two machine learning algorithms, XGBoost and the Adaptive Neuro-Fuzzy Inference System (ANFIS), in predicting heart disease across diverse datasets. Utilizing datasets from the UCI Machine Learning Repository, including Switzerland, Cleveland, Hungarian, Long Beach VA, and Statlog Heart, standard preprocessing techniques such as imputation, standardization, one-hot encoding, and SMOTEENN were applied to ensure consistent modeling conditions. Both models underwent extensive training and optimization. XGBoost excelled, particularly achieving 100% accuracy in the Switzerland and Statlog datasets, while ANFIS demonstrated its strength in modeling complex patterns, notably achieving perfect accuracy in the Cleveland dataset. Performance evaluations using accuracy, precision, recall, F1 score, F2 score, and ROC-AUC score highlighted XGBoost's consistent high precision and recall, vital for reliable CVD diagnosis. In contrast, ANFIS showed potential in clinical settings with its high F2 scores, emphasizing the reduction of false negatives. The study highlights the advantages of using advanced machine learning models like XGBoost and ANFIS in cardiovascular diagnostics, suggesting further research with larger and more varied datasets to refine these models and advance medical diagnostics using machine learning.
İş zekası araçları ile Türkiye'de pazar analizi ve finansal değerlendirmenin geliştirilmesi
The Turkish market emerges as a key player amid global shifts, owing to its unique cultural blend and strategic location. Amid economic intricacies, robust market analysis becomes vital for stakeholders to navigate complexities and seize opportunities. Market research provides a compass for decision-making, enabling identification of trends and mitigation of risks. Business Intelligence (BI) emerges as a transformative tool, blending technology and strategy to harness data for informed decision-making at all organizational levels. As businesses strive for agility and resilience, BI adoption becomes not just a competitive edge but a strategic necessity for sustainable growth in a dynamic landscape. This research provides a thorough method to examine Turkey's market and financial dimensions, utilizing the capabilities of Business Intelligence tools. Its objective is to connect data with practical insights, serving various stakeholders invested in understanding the Turkish economy. Keywords: Macro Economy, Business Intelligence, Financial KPI, Market Analysis, Machine Learning
Makine öğrenme metodolojilerini kullanarak yetkisiz erişime dayanıklı, sağlam bir bilgisayar ağ mimarisi geliştirin
As the use of IT spreads rapidly into new areas, the necessity to ensure the security of these systems has grown. Cyberattacks have also become much more sophisticated as a result of the widespread availability of information technology. Consequently, traditional security measures like SIDS have failed to identify new types of assaults. Intrusion Detection Systems (IDS) make it possible to track and gather harmful data inside a network. The majority of IDSs rely on signatures to identify potential threats. They use a set of rules—either manually entered by the administrator or created automatically by the system—to identify and respond to known threats. To maintain the availability of services at all times, network security experts focus on both preventing and responding to intrusion attempts. To find and categorize suspicious actions, security professionals employ tools like IDS. Hence, to protect privacy, security, and the ongoing delivery of services, it is crucial that the IDS continually keeps up-to-date with the most recent intrusion attack signatures. Important factors to consider while evaluating IDS performance are its speed and its capacity to learn new assaults. This study demonstrates how several Machine Learning techniques may be evaluated using the Knowledge Discovery and Data Mining (KDD) dataset, which is also called Knowledge Discovery in Databases. The primary focus is on creating a comprehensive and representative dataset for experimentation, with a strong emphasis on KDD. For this analysis, we have chosen to use the K-Nearest Neighbor (KNN) and Multilayer Perceptron (MLP) classifiers. The KNN classifier has shown the highest accuracy in recognizing and classifying all types of KDD dataset attacks (DOS, R2L, U2R, NORMAL, and PROBE), both for binary class (NORMAL vs. ABNORMAL) and multi-class scenarios. The experimental findings utilizing the proposed KNN and MLP models showed that the accuracy of binary classification using KNN and MLP was 99% and 97% respectively. Furthermore, the multi-class classification produced improved results compared to earlier work, with reported high-level accuracies of 92% and 87% respectively. This thesis investigates the effectiveness of using deep learning techniques, namely MNN and MLP designs, for network flow-based intrusion detection.
Sağlık IoT cihazlarının güvenliğinin sağlanması ve SIEM entegrasyonunun sağlanması: aAdresleme
This thesis explores the growing use of Internet of Things (IoT) devices in healthcare, which has transformed patient care with advanced monitoring and treatment options. However, this integration poses security challenges, demanding robust protective measures. The study focuses on integrating Security Information and Event Management (SIEM) systems with healthcare IoT devices as a solution. Through a thorough literature review, it sheds light on the current state of IoT security, emphasizing the need for improved protective measures. The main issue addressed is the vulnerability of healthcare IoT devices to security breaches and the associated risks to patient data and device functionality. To address this, the research suggests a comprehensive security framework tailored for these devices. This framework, drawn from literature and best practices, includes crucial security measures such as authentication, data encryption, access controls, and anomaly detection. Integrating SIEM systems into this framework provides real-time threat detection and swift incident response, bolstering the overall security of healthcare IoT devices. The thesis highlights the importance of this integration, offering a roadmap for secure, efficient, and effective implementations of healthcare IoT
Bulut bilgisayar: Bilgi teknolojisi dış kaynak kullanımı için daha iyi bir araç
A demonstrate for giving and utilizing IT administrations online is called cloud computing. This permits businesses to concentrate on their center competencies whereas taking off the complex errands of foundation administration to specialized suppliers. Cloud computing offers an approach to IT outsourcing that is more spry, temperate, and scalable. The appearance of cloud computing has reshaped the scene of data innovation outsourcing, advertising organizations a compelling elective to conventional models. This proposal see into the transformative potential of cloud computing as a predominant implies of outsourcing IT administrations. This consider investigates the benefits that cloud computing offers to the outsourcing worldview by implies of a comprehensive audit of suitable inquire about and case thinks about. It looks at how cloud-based solutions surpass traditional outsourcing contracts in terms of "scalability, cost-effectiveness, accessibility, reliability, security, innovation, and data recovery". Furthermore, this thesis assesses the suggestions of cloud computing on trade procedures, organizational structures, and IT administration hones. It examines the vital contemplations for receiving cloud administrations, including vendor selection, migration strategies, governance systems, and administrative compliance. By synthesizing hypothetical bits of knowledge with observational prove, this study gives a factual understanding of how cloud computing revolutionizes IT outsourcing.
Yüz tanıma sistemleri verimlilik karşılaştırması
Bu çalışmada, yüz tanıma teorilerine ve algoritmalarına artan bir ilgi olduğu belirtiliyor. Video gözetimi, suç tanımlama, bina erişim kontrolü ve insansız araçlar gibi endüstri uygulamalarında bu teknolojilerin önemli rol oynadığı vurgulanıyor. Hayatın birçok alanında kullanıma giren bu yüz tanıma sistemlerinin kendi içinde ayrıldığı yerel yaklaşımlar, bütünsel yaklaşımlar ve hibrit yaklaşımlar adlı sınıfları hakkında detaylı bilgi verilmektedir. Bu teknikler, yüz görüntülerini sadece belirli yüz özellikleri veya tüm yüz özellikleri kullanarak açıklamak için kullanılmaktadır. Sistemler arasındaki verimlilik düzeyleri karşılaştırılmaktadır. Literatür taramasında elde edilen çeşitli araştırma bulguları birbiriyle kıyaslanarak 3 yaklaşım öne sürülen teknikler gözden geçirilmekte ve verimlilik düzeyleri üzerine tartışma sunulmaktadır. Yaklaşımların hayatımızdaki yerleri, yüz tanıma hizmeti sunulurken ortaya sergiledikleri performansı; doğru eşleşmesi, tutarlılığı, sağlamlığı, yapısal durumu, birbirlerinden farkları bakımından avantajları ve dezavantajları listelenmektedir. Görüntü işleme ve bilgisayarlı görme alanındaki çalışmaların hızla yaygınlaştığı son yıllarda, tarım, tıp, eğitim, sağlık ve güvenlik gibi birçok alanda bilgisayarlı görme uygulamaları geliştirilmekte ve günlük yaşantımızda kullanılmaktadır. Özellikle yüz ve nesne tanıma işlemleri, farklı uygulamalarla her geçen gün dahada yaygınlaşmaktadır. Bu yüz tanıma uygulamaları, kullanıcılardan yüz bilgileri toplamak suretiyle veri ve bilgi tabanları oluşturmaktadır. Öne çıkan bir uygulama ise işyerlerinde personel giriş ve çıkışlarını takip etmek amacıyla geliştirilen yüz tanıma tabanlı personel kontrol ve takip sistemidir. Bu sistem, kartlı veya manuel kayıt tutma gibi yöntemlerin yerine daha hızlı, etkili ve doğru bir çözüm sunmaktadır. Giriş ve çıkışlarda kameralar kullanılarak personellerin yüzleri tespit edilir ve kimlik doğrulaması yapılır. Bu sayede personellerin işe geliş, gidiş saatleri ve fazla mesai bilgileri otomatik olarak takip edilebilir hale gelir. Geliştirilen bu sistem, özellikle hijyen ve sağlık koşulları açısından kart veya parmak izi gibi geleneksel yöntemlerin zorlayıcı olduğu durumlarda etkili bir çözüm sunmaktadır. Biyometri, bireyleri birbirinden ayırt etmeyi mümkün kılan fiziksel ve davranışsal özellikleri inceleyen bir bilim dalıdır. Bu kapsamda, biyometrik sistemler, bireylerin kimliklerini belirlemek amacıyla özel biyometrik özellikleri kullanarak tasarlanmış sistemlerdir. Bu sistemler arasında yer alan yüz tanıma sistemleri, bireyleri tanımlamak için yüz özelliklerini kullanır. Yüz tanıma sistemleri genellikle güvenlik ve personel devam kontrolü gibi alanlarda, cep telefonlarında, sosyal medyada, bina giriş ve kontrol alanlarında yaygın olarak kullanılmaktadır. Bu çalışmanın amacı mevcut yüz tanıma sistemlerinin uygulama alanları hakkında bilgi vermek, yüz tanıma sistemlerinin sınıflarını incelemek ve verimliliklerini karşılaştırmaktır. Anahtar Kelimeler: Yüz Tanıma Sistemleri, Güvenlik Sistemleri, Biyometrik Sistemler, Kişi Tanımlama, Verimlilik
Şehir içi trafik sıkışıklığını azaltmak için bir araba paylaşım uygulaması tasarımı
Urban areas across the globe are negatively impacted by traffic congestion, and Istanbul is no exception. The city's traffic situation is especially daunting due to its large population of 16 million individuals who reside and work there [1]. To address this issue, carpooling has been identified as an effective method for decreasing traffic congestion and promoting sustainable transportation. As a result, we created a carpooling application tailored for Istanbul with the goal of mitigating traffic congestion. The app permits drivers and passengers to generate personal profiles, seek out carpooling companions, and communicate with one another to coordinate rides. In addition, the app offers services like booking rides, processing payments, and assessing and evaluating experiences to guarantee safety and liability. Overall, to ensure the success of the project, the app will be designed to address the specific needs of Istanbul's urban population and to provide a viable solution to the city's traffic congestion problem using the database schema and API architecture using the PHP programming language and MySQL database management system. The app will be also scalable and can be adapted to other urban areas facing similar traffic problems and compatible with multiple platforms.
Derin öğrenme aracılığıyla hibrit tespit geliştirmeleriyle ağ ve sunucu giriş tespiti
Network security and the identification of possible threats depend heavily on anomaly detection in network traffic. In order to determine which machine learning models are most useful in detecting network traffic anomalies, this research compared and evaluated a number of different machine learning models. Deep Learning with 10 epochs, Logistic Regression, Random Forest with a Filter method, Random Forest with a Wrapper method, and Random Forest with a Wrapper method were the models evaluated. A number of metrics were used for evaluation, including cross-validation, accuracy, F1-Score, ROC (Receiver Operating Characteristic), and precision- recall. The results showed that the Random Forest with the Wrapper approach and the Deep Learning model outperformed other assessment criteria. With a ROC score of 0.9716, the Deep Learning model had the best performance, clearly demonstrating its superior ability to differentiate between regular and abnormal network traffic. It also displayed outstanding precision-recall scores of 0.9621, indicating accurate anomaly identification with a low incidence of false positives. The Deep Learning model also attained a strong F1-Score of 0.8405, which represents a balanced mix of recall and precision.
Siber güvenliği geliştirmek dijital görüntülerle oluşan siber tehditlerin azaltılması
This thesis introduces a cutting-edge Virus Detection App designed to address cyber threats embedded in digital images, with a particular focus on the underlying algorithms, specifically leveraging artificial intelligence (AI). The core of the application revolves around the implementation of the InceptionV3 deep learning model, a powerful convolutional neural network (CNN) renowned for its image recognition capabilities. The algorithmic workflow begins with the preprocessing of digital images. Images are resized to a standardized 300x300 pixel format to insure thickness in the input data. The preprocessing step is pivotal for optimizing the model's performance by furnishing invariant input data for analysis. Transfer literacy is a crucial element of the algorithmic approach. The InceptionV3 model,pre-trained on the ImageNet dataset, serves as a point extractor. By using the knowledge acquired during its expansive training on a different set of images, the model can discern intricate patterns and features in the digital images applicable to malware, contagions, and trojans. This transfer of knowledge significantly accelerates the training process and enhances the model's capability to generalize across different image datasets. The comprehensive scanning point, a highlight of the operation, involves the methodical operation of the trained model to a stoner- named set of images. The model's prognostications are decrypted, and implicit pitfalls are linked. The choice of a top- 3 vaticination strategy offers a nuanced understanding of the model's confidence in its assessments. Continual literacy is eased through a stoner- touched off model training process. The model is streamlined on a dataset, icing its rigidity to evolving cyber pitfalls. The use of a thick subcaste with a double sigmoid activation function enhances the model's perceptivity to relating infected images while minimizing false cons. This exploration underscores the critical part of sophisticated algorithms, embedded in artificial intelligence, in creating a robust and adaptable Contagion Discovery App. The admixture of preprocessing, transfer literacy, and continual training contributes to the efficacity of the operation in relating and mollifying cyber pitfalls in digital images.
Üstün pnömoni tespiti için bok böceği ve fick yasası kullanılarak hibrit optimizasyonla geliştirilmiş yeni bir derin öğrenme çerçevesi
Pneumonia is a leading cause of global mortality, particularly among children and the elderly, and is characterized by inflammation of the air sacs in the lungs. Early and accurate detection is crucial to mitigating its spread and improving patient outcomes. Traditional diagnostic methods, including physical exams, blood tests, and chest X-ray analysis by radiologists, are effective but often time-intensive, resource-heavy, and prone to human error. These limitations underscore the need for advanced, efficient diagnostic solutions. This study leverages deep learning techniques, specifically Convolutional Neural Networks (CNN) and MobileNet architectures, to enhance pneumonia detection using chest X-ray images. Deep learning excels in image recognition tasks through automatic feature extraction, making it a powerful tool for medical diagnostics. The research applies these models to a large dataset of chest X-rays labeled as NORMAL or PNEUMONIA, utilizing rigorous preprocessing techniques, including normalization and augmentation, to improve data quality and diversity. Hybrid optimization techniques inspired by the dung beetle optimizer and Fick's law of diffusion are employed to optimise model performance, effectively navigating complex parameter spaces. Evaluation metrics such as accuracy, precision, recall, and F1-score validate the models. Results demonstrate the enhanced MobileNet model's superior performance, achieving an accuracy of 98.19%, significantly outperforming the CNN baseline. This highlights the potential of deep learning and innovative optimization methods in advancing pneumonia diagnostics. The findings emphasize the broader implications of artificial intelligence in healthcare, offering a pathway for developing robust diagnostic tools that expedite accurate and timely disease detection. This research contributes to alleviating the global pneumonia burden and underscores the need for continued innovation in machine learning to enhance medical outcomes.
Adaptive decision tree with random forest integration and dimensionality reduction for efficient botnet forensics
Adaptive Decision Tree with Random Forest Integration and Dimensionality Reduction for Efficient Botnet Forensics proposes a new method for botnet detection by combining an adaptive decision tree with random forest integration. As response variables are highly dimensioned in botnet detection, multistep and time-consuming detection processes are major challenges. With an integrated method, we could first model the response variables as multiclass and regression formats to simplify the construction of decision trees. Then, based on the adaptive decision tree feature selection, we filter out non-obvious features to efficiently establish the random forest regression model under the appropriately sized feature space. Furthermore, to handle multilabel classification, a random forest is employed as the global model in Tree-2-Rule procedures to detect botnet-affected communication behaviors. Finally, real data experiments have been conducted based on the top datasets. The results show that the adaptive decision tree has excellent improvements in efficiency and accuracy. In future research, the GPU ninth-ordinal censored multistate diagnosis data is useful observed materials that do not destroy the random effect influence of the data. Also, whether the application of the random forest model could save more time in analyzing existing commercial status is an issue to be clarified in future development. Additionally, the development of the method proposed in this research requires further investigation. We could improve and then propose the preferable solution based on the research results. Our proposed solution can be utilized as a tool for efficient real-time bot incident investigations, in accordance with both academic and business objectives.
Derin öğrenme modeli ile zatürre tespiti
Zatürre, her bireyin yaşamının herhangi bir döneminde maruz kalabileceği bir akciğer enfeksiyonudur. Tedavi edilmediği takdirde ciddi sağlık sorunlarına ve hatta ölüme neden olabilir. Bu nedenle, hastalığın erken teşhisi büyük bir öneme sahiptir. Zatürrenin tespitinde çeşitli yöntemler kullanılmakla birlikte, en yaygın yöntemlerden biri akciğer röntgen görüntülerinin incelenmesidir. Bu görüntüler uzmanlar tarafından titizlikle değerlendirilir. Ancak bu sürecin doğruluğu, süresi ve harcanan emek önemli faktörler arasında yer almaktadır. Görüntü analizi alanında, yapay zekanın hızla gelişmesiyle birlikte önemli çalışmalar yapılmıştır. Bu tezde, akciğer röntgen görüntüleri kullanılarak makine öğrenmesi ile zatürre teşhisini destekleyen bir yöntem geliştirilecektir. Böylece, teşhis sürecinde zaman yönetimi, doğruluk oranı ve harcanan çabanın optimize edilmesi sağlanacaktır. Literatürdeki çalışmalar ve geliştirilen modeller dikkate alınarak, evrişimsel sinir ağı tabanlı yöntemler incelenecektir. Çalışmada kullanılan akciğer görüntüleri üç gruba ayrılmıştır: eğitim, doğrulama ve test. Sonuç olarak, evrişimsel sinir ağı modellerinin kullanılmasıyla elde edilen sonuçların, aynı konu üzerinde araştırma yapan araştırmacıların sonuçları kıyaslanarak literatüre katkıda bulunması ve doktorlara karar desteği sağlaması amaçlanmaktadır.
Exploring social influence in the design and evaluation of enhanced password meters with peer-driven security
Passwords have controlled the world of authentication. As a result of their extensive use, they have become a desired target for attackers. To thwart such assaults, many strategies have been used to increase password security. The websites and applications use password meters to assist users in creating complex passwords. The purpose of a password meter is to give the users knowledge about their password selection by identifying it as "weak," "medium," or "strong," for example. This study has taken into account social effect, or how the other influences a person's behavior and attitude. This social influence, often known as peer feedback, was taken into account while designing a peer feedback password meter. Participants who used the peer feedback meter and were instructed to create a special password had stronger passwords than those who used the conventional meter. In this research, many works related to methods of measuring password strength were reviewed, and algorithms used to measure password strength were also reviewed. Some of the ways in which the attacker hacked passwords were also mentioned. We aim to provide effective methods for evaluating passwords with five levels of integrity.
Towards cleaner environments: A study of deep learning models for plastic bottle detection
The widespread use of plastic bottles in our daily lives is contributing to significant environmental issues, particularly in marine ecosystems. A substantial quantity of plastic bottles is being carried back to the mainland from the sea by waves, often becoming trapped in coastal areas. The detrimental impact of plastic waste, including plastic bottles, on coastal ecology is a matter of concern. Thankfully, artificial intelligence (AI) has found diverse applications in various sectors, including environmental initiatives, offering promising solutions to combat these environmental challenges effectively. This report aims to provide a classification of bottle plastic using data images in various situations. We preprocess with the dataset and we apply different artificial intelligence algorithms to perform this classification. The CNN model achieved the highest ACC, reaching an impressive 99%.
Ethical implications of ai driven personalization in digital marketing
The digital marketing industry is undergoing a transformation owing to the use of machine learning with Artificial intelligence (AI). This technology permits the industry to provide a highly personalized experience for consumers. As soon as it derives to AI, there's considerable discourse about engagement and devotion. The purpose is simple: AI bids a way to distinguish "stuff" at rate never before conceivable. Hitherto, despite the advent of these glowing outcomes, numerous ethical queries need answering. These ethical features have been inspected through thorough study of the literature and case studies to attain a modernized sorting of noticeable issues. These are: (1) the irresistible quantity of data being gathered and the confidentiality risks required; (2) the prejudices that algorithms can symbolize and the discrimination they can preserve; (3) the likely for consumer influence; (4) the economic disruption that AI might cause; (5) the transparency issues and the accountability shortfall that it generates. Although AI-powered personalization is able to produce additional pertinent and tempting consumer familiarities, it also presents serious ethical predicaments that must be addressed to guarantee responsible and evenhanded applications. The study concludes through recommendations for vendors and legislators to resolve the remunerations of AI personalization with the inevitability for ethical morals and consumer fortification, ultimately aiming to respect human principles and improve inclusive well-being.
Rastgele Orman'ın ötesinde: Yorumlanabilirlik gereksinimleri ile sigorta satın alma tahmini için gelişmiş topluluk ve hibrit modeller
This study focuses on creating machine learning models to predict whether people will buy insurance. We developed and tested sophisticated ensemble models that aim to be both highly accurate in their predictions and easy to understand: an important balance since insurance companies need to explain their decision-making processes to meet regulatory requirements. Using a dataset of 53,503 customer records, a comprehensive data preparation pipeline transformed 20 raw features into 100 engineered variables, thereby capturing temporal, behavioural, and categorical patterns. Class imbalance was addressed using Synthetic Minority Oversampling Technique (SMOTE) which ensures equitable model learning. Models which include Random Forest, Support Vector Machine (SVM), LightGBM, Graph Neural Networks (GNNs), and hybrid designs were compared, with Random Forest and LightGBM achieving near-perfect performance (F1-score: 1.0000 and 0.9999, respectively). GNNs, integrated by means of a k-NN similarity graph, showed scalability but small performance gains due to the sufficiency of original features. Since interpretability was a focus of this research, SHAP and LIME were used and their use revealed that temporal features such as "Days_Since_Purchase" and "Purchase_Year" were the main contributing factors. Also, optimization with multiple objectives using Optuna achieved a Pareto-optimal configuration (ROC-AUC: 1.0, F1-score: 0.9896, fidelity: 0.9843, sparsity: 3.0) which satisfies strict regulatory requirements (GDPR, NAIC). A case study on 1,000 customers showed the trade-off between high-performance LightGBM+GNN (ROC-AUC: 0.6211) and interpretable Decision Trees (ROC-AUC: 0.5380), with stakeholder feedback favouring transparency. The study recommends making temporal and monetary features priorities for customer segmentation and examining time-evolving GNN architectures and alternative graph constructions for future research to improve predictive power in noisier datasets. This framework provides a scalable, interpretable solution for insurance analytics, and that supports targeted marketing and compliance.
Eliminate entanglement in quantum information processing utilizing hybrid quantum-classical neural networks
Quantum entanglement is a physical phenomenon that lies at the heart of the contrast between quantum physics and classical physics, which distinguishes quantum mechanics as an essential feature. It happens when a collection of particles interacts with one another, participates, or creates a collection of particles in close spatial proximity in a dependent way in which the quantum state of one particle in this collection cannot be characterized independently of the other particles' quantum states. The states of quantum mechanics have several types of order, including Symmetry Protected Topology (SPT), in which matter has symmetry and a finite energy gap at zero temperature. Feedback set methods are used to derive results (how to identify specific quantum data source attributes) in a more consistent manner. The concept of quantum entanglement refers to that SPT states are symmetric short-range entangled states., since only "trivial" topological orders characterize short-range entangled states. The researcher presents and analyzes: Convolutional neural networks inspired by a quantum circuit-based technique, which is highly effective when applied within a hybrid neural network environment allowing them to be efficiently trained and executed on realistic quantum devices, this research simplifies and implements a quantum convolutional neural network (QCNN) in a hybrid neural network environment on the TensorFlow platform, where it functions as a suggested quantum equivalent to a conventional convolutional neural network. This thesis demonstrates how to detect certain characteristics of a quantum data source, such as a device's intricate simulation or a quantum sensor. With or without an excitation, a cluster state will operate as the quantum data source, which the QCNN will learn to identify. In this study, the researcher classed the SPT phase as a dataset. The QCNN system attained a validation accuracy of 100% Instead of 89.58% in its purely quantitative analog, we use three layers of quantum convolution followed by a classical densely connected neural network to identify an exciting cluster state within 10 epochs of time. This architecture should be quite successful in mitigating entanglement.
Hisse senetlerindeki trendlerindeki dalgalanmaları öngörmekderin öğrenme
Recent studies indicate that variations in the value of the stock market are difficult to anticipate due to the large number of unknowns and variables that affect its value on any given day. This includes the current market conditions, investor opinion toward a certain firm, and political developments. Hurriedly and arbitrarily, the pricing marketplaces are selected when setting the stock price. Due to the fact that it is not rare for the stock market to be dynamic and disorderly (due to several reasons), the stock market's direction is categorized as a random process, with more shifting possibilities in short time frames Therefore, an attempt to carry out a price forecast in the stock market can bring great benefits to investors, by increasing the level of information about the financial market, minimizing exposure to financial risk. In this sense, a computational technique called Artificial Neural Networks (ANNs) can be applied. The Artificial Neural Network (ANN) simulates on computers the functioning of the human brain in a simplified way.