40
Archived Theses
0
DOIs Assigned
0%
DOI Rate
Discipline
Makine öğrenimi görevleri için modern veri yönetim stratejileri: Bulut platformu üzerinde spor analitiği uygulaması
There is no doubt that data is the most valuable asset today. The efforts of enterprises in digital transformation and creating a data-driven culture are the most concrete indicators of this. Nowadays, where data transforms all industries, it is possible to follow the rapidly developing technological developments in this field. Appropriate data management strategies are the basis of creating data-driven organizations. When the evolution of data management architectures is examined, it is possible to say that the biggest factor that triggers this evolution is the changing and increasing data sources and the velocity of data production. In addition, with the increase in the importance of business use cases that need to be done in real time, it has become a very crucial need to process data quickly and turn it into action. Today when data is strategic importance, enterprises that can manage data correctly could gain competitive advantages. Being able to the correct data management can be built with the support of up-to-date and modern approaches. The infrastructures established by the integration of new and modern methods into the platforms turn into more agile structures. This increases the number of value added services to be produced from data by providing speed and flexibility to organizations. Today, the outputs expected to be produced from data management platforms go beyond descriptive and diagnostic analytic. Now, artificial intelligence, machine learning and data science are important parts of these platforms and these are opened new channels for the future of enterprises. In this thesis, basic needs and capabilities of modern data management architectures are described and detailed explanations were made on reference architectures in the industry. Besides, data management strategies and expectations were discussed. An example prototype of the data management platforms, which is explained in detail in this thesis, has also been developed on the cloud platforms. In this prototype, the entire life cycle of the data was considered and each step was developed in detail. In addition, a data science project was developed using the data collected on the platform. Thus, an end-to-end solution has been implemented.
Hisse senetlerinde işlem otomasyonu için derin güçlendirme öğrenme yaklaşımı
Deep Reinforcement Learning (DRL) algorithms can scale to previously intractable problems. The automation of profit generation in the stock market is possible using DRL, by combining the financial assets price "prediction" step and the "allocation" step of the portfolio in one unified process to produce fully autonomous system capable of interacting with its environment to make optimal decisions through trial and error. In this study, a continuous action space approach is adopted to give the trading agent the ability to gradually adjust the portfolio's positions with each time step (dynamically re-allocate investments), resulting in better agent-environment interaction and faster convergence of the learning process. In addition, the approach supports the managing of a portfolio with several assets instead of a single one. This work represents a novel DRL model to generate profitable trades in the stock market, effectively overcoming the limitations of supervised learning approaches. We formulate the trading problem as a Partially Observed Markov Decision Process (POMDP) model, considering the constraints imposed by the stock market, such as liquidity and transaction costs. More specifically, we design an environment that simulates the real-world trading process by augmenting the state representation with ten different technical indicators and sentiment analysis of news articles for each stock. We then solve the formulated POMDP problem using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm and achieved a 2.68 Sharpe ratio on the test dataset. From the point of view of stock market forecasting and the intelligent decision-making mechanism, this study demonstrates the superiority of deep reinforcement learning in financial markets over other types of machine learning such as supervised learning and proves the credibility and advantages of strategic decision-making using DRL.
Bir parti büyüklüğü oyununda özgün makine öğrenmesi tekniklerini kullanarak maliyetlerin dağıtılması
In supply chain management (SCM), effective resource utilization is the key to achieving certain strategic benefits such as minimizing costs, increasing service levels, reducing inventories, increasing responsiveness, and finally improving customer satisfaction. Collaborative approaches among supply chain entities have become increasingly popular to increase resource utilization. In this thesis, we analyze a collaborative production setting where several companies facing varying demands throughout a finite planning horizon attempt to reduce their procurement costs by ordering from a common supplier. As the capacity of the common supplier is better utilized in such a collaborative solution, it will yield benefits that will be shared by the collaborators. Our objective is to design a cost allocation framework to ensure the sustainability of the collaborative purchasing organization. We propose various methods, including novel and computationally efficient machine learning based methods, using two different architectures, gradient boosting mechanism, and artificial neural networks which ensure the scalability of the proposed framework. We perform an extensive computational study and observe that our proposed method significantly outperforms the generic methods in the literature in terms of solution quality and computation time.
Müşteri şikayetleri üzerinden doğal dil işleme teknikleri ile beyaz eşyalarda arıza teşhisi
The thesis investigates the application of Natural Language Processing (NLP) for fault classification in the white goods sector, using real-world service data. two-stage hybrid pipeline is developed: first, unstructured customer complaint texts are transformed into structured representations through both rule-based preprocessing and Large Language Model (LLM)-based enrichment. Then, multiple machine learning classifiers are trained to predict repair codes, which include the Random Forest, Naive Bayes, LightGBM, BiLSTM CNN-BiLSTM and DistilBERT. The system is evaluated on a large-scale dataset of 103,000 technical service records. Experimental results reveal that semantically enriched representations generated by GPT-4o-mini consistently outperform both unprocessed and rule-based inputs. Across all classifiers, the LLM-Rich preprocessing strategy yields the highest performance, achieving complexity-weighted average F1 scores of up to 63,4\%, compared to 54,1\% for the best unprocessed input and 52,2\% for the best rule-based alternative. Furthermore, the study introduces a complexity score for each repair code, allowing analysis of model performance relative to task difficulty. Results indicate that LLM-enhanced inputs not only improve accuracy but also increase robustness across high-complexity classes. Overall, the results indicate that semantically enriched NLP pipelines offer a scalable and powerful basis for semi-automated fault diagnosis, particularly in low-resource languages like Turkish, and greatly strengthen the robustness of the corresponding classification models.
Yapay sinir ağları ile çok etiketli sınıflandırma
Multi-label classification has huge importance for several applications, it is also a challenging research topic. It is a kind of supervised learning that contains binary targets. The distance between multilabel and binary classification is having more than one class in multilabel classification problems. Features can belong to one class or many classes. There exists a wide range of applications for multi-label prediction such as image labeling, text categorization, gene functionality. Even though features are classified in many classes, they may not always be properly classified. There are many ensemble methods for classification. However, most of the researchers have been concerned about better multi-label methods. Especially little ones focus on both efficiency of classifiers and pairwise relationships at the same time to implement better multi-label classification. In this paper, we worked on modified ensemble methods by getting benefits from k-Nearest Neighbors and neural network structure sequentially to address issues beneficially and to get better impacts from the multi-label classification. Publicly available datasets (yeast, emotion, scene, and birds) are performed to demonstrate the developed algorithm efficiency, and the technique is measured. Our algorithm outperforms benchmarks for each dataset with different metrics. The result of the algorithm is competitive with the state-of-the-art results. Especially, in the weighted average of false-positive minimization and false-negative minimization, the algorithm passes the benchmarks.
Kanallar arası dolandırıcılık tespiti için yeni bir örnekleme tekniği ve gradyan artırıcı ağaç tabanlı yaklaşım
The most recent research on hundreds of financial institutions uncovered that only 26% of them have a team assigned to detect cross-channel fraud. Due to the developing technologies, various fraud techniques have emerged and increased in digital environments. Fraud directly affects customer satisfaction. For instance, only in the UK, the total loss of fraud transactions was £1.26 billion in 2020. In this study, we come up with a Gradient Boosting Tree (GBT)-based approach to efficiently detect cross-channel frauds. As a part of our proposed approach, we developed an algorithm able to generate an optimized training set to train the model and overcome imbalanced data problems. This solution made it easier for the model to understand the concept drift, another major problem arising from changing customer behavior. We boost the performance of our GBT model by integrating additional demographic, economic, and behavioral features as a part of feature engineering. Hyperparameter tuning methods find the best parameters for the model. The cross-channel fraud detection performance of the model is evaluated on a real banking dataset which is highly imbalanced in terms of fraud which is another challenge in the fraud detection problem. We use our trained model to score real-time cross-channel transactions by a leading private bank in Turkey. As a result, our approach can catch almost 75% of total fraud loss in a month with a low false-positive rate and acceptable call count.
Deterministik doğrusal eşik modelinde n-hop etki maksimizasyon problemi
The Influence Maximization Problem (IMP) finds a set of highly influential nodes within a social network in order to maximize the spread of influence. We consider that people can have influence on their direct (1-hop), 2-hop and 3-hop neighbors. IMP with extended influence transitivity is called n-hop IMP. In this paper, we study the problem under the deterministic linear threshold model and propose a heuristic solution. In our proposed heuristic model, there are two parts, extended seed set algorithm and local search. Main purpose of extended seed set algorithm is creating a node set and send it to local search which is based on trying to improve it via replacing the nodes in the given set. We used two different features for selecting the candidate nodes. We propose an equation to estimate the value of a node set without actually computing. We used real-life and synthetic networks to test our solution method and generated weight and threshold values in three different methods.
Gelişigüzel düğüm dağılımlarına sahip GSP örneklerinin en iyi tur uzunluğunu tahminlemek için sinir ağı tahminleyicileri
To achieve operational efficiency in logistics, we need to solve complex routing problems. Due to their complexity, these problems are often solved sequentially, i.e., using cluster-first route-second (CFRS) type frameworks. However, such two-phase frameworks generally suffer from sub-optimality arising from the first phase. To mitigate this sub-optimality, information about optimal tour lengths of potential clusters can be exploited first, thereby transforming this two-phase approach into a less myopic solution framework. In that aspect, a quick and highly accurate Traveling Salesperson Problem (TSP) tour length estimator can be utilized for searching high-quality clusters. Motivated by this, we propose novel and computationally efficient neural network-based optimal TSP tour length estimators. Our approach uses an entirely new feature set consisting of node level, instance level, and solution level features by combining the power of artificial neural networks and theoretical knowledge in the routing domain. This data and knowledge hybridization enables us to achieve predictions with less than 0.7 percent deviation (on average) from the optimality. Unlike previous studies, we design and use new instances mimicking real-life logistics networks and morphologies. These instance characteristics introduce a substantial computational cost, making our instances harder to solve. To cope with these pathologies, we devise a new and efficient way of finding lower bounds and partial solutions to TSP later to be used as solution-level predictors. We also conduct a computational study where we produce up to 100 times lower prediction error on out-of-distribution test instances. Finally, we develop an enumeration-like mechanism by incorporating proposed machine learning models and metaheuristics to solve massive-scale rout- ing problems efficiently. We significantly outperform the state-of-the-art solver in terms of solution time and quality, demonstrating the potential of our models and the proposed method.
Machine learning based allocation in a lot sizing game
Supply chains often have confliction objectives and operate in a finite resource setting. Traditionally, companies focused on their internal processes to generate cost reduction opportunities in order to increase their profitability. However, recent studies suggest that collaboration is the key to have sustained benefits among supply chain partners. Supply chains, which usually have conflicting objectives, can form coalitions and take advantage of collective payoffs. In this paper, we analyze a collaborative production setting where several companies facing varying demands throughout a finite planning horizon attempt to reduce their procurement costs by ordering from a common supplier. The participants exploit the synergy among themselves by maximizing the capacity utilization of the supplier. We design a novel cost allocation method using various machine learning techniques with the goal of generating a cost-allocation mechanism that ensures the sustainability of the collaboration. We conduct a computational study to compare and contrast our proposed method with the generic methods in the literature. We discuss the advantages and disadvantages of these methods in terms of solution quality and computation time.
Machine learning framework to price-setting risk-averse data-driven newsvendor problem
Many business-owners, including manufacturers, retailers, distributors are in need to have decisions to make ordering for tons of products in their daily routines. This decision-making cannot be considered trivial since there is always a risk of ordering more or less then what is actually required, which may result in severe undesirable circumstances that are overstocking or shortages. The newsvendor problem in the literature focuses on this trade-off in its classical applications. In such approaches, there is an assumption of demand distribution is known. However, true demand is almost never known and constantly change due to being open to oodles of internal and external parameters; therefore, it is almost impossible to be known and hard to be predicted in real life. For this reason, traditional approaches to newsvendor model in the literature are very valuable yet it is possible to create different variations of this problem which are more appropriate for the real world instances. In addition to quantity of ordering, the determination of price is another serious concern of decision-makers. In order to satisfy the need of determining correct amount of order and regarding price positioning, there is a different approach to newsvendor problems, which is price-setting providing not only quantity but also price information. In this paper, we focus on this variant of newsvendor problem and create a data-driven price-setting newsvendor model, where demand is completely unknown. What is unique in this work is to integrate another element in addition to price decision, which is enabling a risk-averse decision-making perspective. In this way, the constructed model decides on quantity of order and price for products, as well as satisfies a certain risk constraint in order to a set a barrier of probability to not generate under a predetermined level of profit. The price-setting and risk-averse applications of newsvendor model exist in the literature yet combining these two different perspectives is what makes our approach novel.
NFT pazarının tanımlayıcı ve tahmine dayalı analizi
Non-fungible tokens (NFTs) are digital assets on a blockchain that have unique identi- fication codes and metadata that make them distinguishable from one another. NFTs can represent a wide range of digital assets, including game cards, artwork, and even real estate. Due to these characteristics, NFTs have gained a tremendous interest from people around the world, leading to huge returns on investment in the NFT market. However, there are only a few studies on the market in the literature. This paper examines various aspects of the NFT market to shed light on its dy- namics and wallet behaviors. First, a descriptive analysis of the market is performed to show its overall trend. The transactional behaviors of wallets are then analyzed, and a segmentation is made to gain a general understanding of the user portfolio. The buyers of a specific NFT collection (Bored Ape Yacht Club) are then studied by comparing them to the overall market, revealing differences in transactional tenden- cies and macro indicators. Finally, machine learning models are developed to predict the transactional behaviors of wallets. Our analysis has revealed that the growth of the NFT market is largely driven by new entrants to the market, but lately there has been a significant decrease in the number of new wallets entering the market. We have also found that the majority of wallets in the market have only one transaction and hold only one token, suggesting that these are users who are experimenting with the market. When we look at the Bored Ape Yacht Club sample, however, we see that these users are highly engaged with the market, with high trading frequencies and a diverse portfolio. Finally, our predictive models show that the transactional behaviors of wallets can be predicted, which opens up opportunities for optimization in various areas.
Makine öğrenimi algoritmalarını kullanarak buzdolaplarında otomatik arıza tespiti
The sustainable functioning of refrigerators is crucial in residential and commercial settings. These appliances are used continuously throughout the day, and any failure can lead to food spoilage, negatively impacting brand reputation. Therefore, having an efficient failure detection system that can identify and diagnose any problems instantly is essential. This paper proposes a novel machine learning pipeline that uses online sensor data from the refrigerators of anonymous customers and a feedback mechanism to inform customer service about the detected failure remotely. The performance of the system is evaluated through a real-life pilot project, and the results indicate that the proposed method achieves high accuracy in detecting various types of failure. Applying the proposed approach prevents food spoilage, reduces maintenance costs while increasing customer satisfaction, and enhances the reliability and safety of refrigerators.
Resim işleme ile kripto varlıkların gelecekteki fiyat hareketinin tahmin edilmesi
Digital or virtual currency known as cryptocurrency uses cryptography for security and is not controlled by a central bank. Cryptocurrencies control the issue of new units and record transactions using decentralized technology, such as blockchain. Cryptocurrencies are entirely digital and have no physical form, in contrast to traditional currency, which is real and backed by a government or financial institution. Although Bitcoin was the first and best-known cryptocurrency, there are now thousands of other coins in use, including Ethereum, Tether, BNB, XRP etc. Bitcoin's value can be extremely unstable and it is frequently utilized as an investment or speculative asset. In some locations, it can also be used to make purchases of products and services, and some companies even accept it as payment. Due to the formation of this sector in relation to the growth in earnings and followers, all focus turned in the direction of the cryptocurrency market. Despite the abundance of studies that have been done in the past for this area, less image processing research has been done for the next movement prediction. This paper uses Bitcoin (BTC) dataset and tries to create a tool to project the upcoming price direction. The target variable is a binary type as the next movement will decrease or increase direction. The challenge of forecasting the next day's price movement involves learning from the information from the previous day by transforming them to the images. Since the main goal is using image processing for prediction, Convolutional Neural Networks, one of the most well-known deep learning techniques, and human judgment will be used. As a result, it is aimed to create an algorithm that makes a successful buy-sell decision on BTC and to achieve profitability. However, if the algorithm created for this study has adequate performance and structure on BTC, it will be simple to adapt it to other cryptocurrency kinds in the future.
Dynamic capacitated lot sizing problem optimization with reinforcement learning
In today's fast-paced manufacturing and supply chain landscapes, efficiently managing production quantities amidst capacity constraints and demand fluctuations is crucial. This MSc. thesis explores a cutting-edge solution to this challenge by harnessing the power of Reinforcement Learning (RL). Reinventing traditional optimization approaches, RL offers adaptability and intelligence to address the Dynamic Capacitated Lot Sizing Problem (DCLSP). Through extensive experimentation and comparison with conventional methods, our RL-based framework showcases superior performance, reducing costs and enhancing production efficiency. This research unlocks a new era of agile decision-making in complex manufacturing environments and opens doors to further advancements in operations management.
NFT alım-satımında veriye dayalı bir yaklaşım: Q-öğrenme tabanlı simülatör
Non-fungible tokens (NFTs) have garnered considerable attention in recent years due to their broad range of applications and potential as a lucrative investment opportunity. Given the nascent nature of the technology and the scarcity of comprehensive studies, there is a pressing need for a holistic trading framework that not only addresses the complexities inherent in the trading process but also proposes viable solutions to existing challenges. This study introduces a novel approach for navigating the NFT trading landscape, effectively confronting the various challenges and suggesting practical solutions. The trading environment is modeled using a Markov Decision Process (MDP), with Q-learning employed to simulate the environment and resolve the MDP problem. The study proposes machine learning models to tackle key challenges, including defining the market state, appraising NFT tokens, and addressing the illiquidity issue prevalent in the NFT market. The proposed approach yields an NFT trading strategy that has shown to outperform traditional strategies, generating substantial profits even amidst bearish market conditions. The Bored Ape Yacht Club (BAYC) collection serves as the primary data set, with the agent trained from June 1, 2021, through January 1, 2023. In testing period from January 1 to June 10, 2023, the proposed model outperformed traditional benchmarks, achieving a profit of 21.14% as opposed to a 20.39% loss for the best-performing benchmark. We assert that this forms a robust foundation for future research into NFT trading simulation and backtesting. We also identify potential areas for future enhancements, particularly possible improvements in the trading strategy and Q-learning approach. The insights gleaned significantly enhance the understanding of the importance of AI applications in the rapidly evolving field of NFT trading.
A revised approach to cryptocurrency portfolio optimization using advanced Q-learning and policy iteration frameworks
Despite all the factors that cause concern among investors, such as volatility and decentralization of crypto world, the popularity of cryptocurrencies continues to grow steadily. The cryptocurrency market still holds its allure for many investors due to the high profit levels it has experienced in the past. With the entrance of numerous altcoins into the market, portfolio management becomes much more challenging. In the literature, we come across numerous studies proposing efficient portfolio management techniques for cryptocurrencies. This study presents proposed models developed based on policy iteration and Q-learning algorithms. Under Q-learning, three distinct sub-models are introduced: Deep Q-Network (DQN), Double Deep Q-Network (DDQN), and Double Dueling Q-Network (DDDQN). All of these models are trained using 6-month training periods and compared using 10 different training and testing periods. Additionally, to evaluate both of proposed policy iteration and Q-learning models, baseline models were created for each algorithm, and the performance of the proposed models was assessed against these baseline models. The results indicate that among Policy Iteration models, the proposed model has the highest average ROI value of 3%, making it the top-performing model. Similarly, among Q-learning models, the proposed DQN model surpasses both baseline models and other Q-learning models, with an average ROI value of 2%. Considering all the models, the proposed Policy Iteration model achieves the highest average ROI value, while the proposed DQN and the proposed DDDQN model demonstrates the lowest volatility in terms of ROI standard deviations.
Envanter rotalama optimizasyonunda entegre bir makine öğrenmesi çözüm yaklaşımı
Inventory Routing Problem (IRP) arises from vendor-managed inventory business set- tings where the supplier is responsible for replenishing the inventories of its customers over a planning horizon. In the IRP, the supplier makes the routing and inventory decisions together to improve the overall performance of the system. In our setting, the supplier's goal is to minimize total transportation costs over a planning horizon while avoiding stock-outs at the customer locations. We assume that the supplier has a fleet of homogeneous capacitated delivery vehicles and abundant availability of the product to be delivered to the customers. Each customer has a constant de- mand/consumption rate and limited storage capacity to keep inventory. To address this problem, we propose a novel integrated clustering and routing algorithm. In the clustering phase, we partition the customer set into clusters, ensuring that each cluster is served by a single vehicle. To accomplish this, we employ a novel deep learning model within the clustering framework. In the routing phase, we develop the delivery schedule for each cluster. What sets our approach apart is its consider- ation of the three key decisions—when to deliver, how much to deliver, and how to route—by integrating both a mathematical model and a machine learning model in the decision-making process. We evaluate the performance of the proposed clustering and routing algorithms against existing literature, and our results demonstrate sig- nificant improvements. Furthermore, the proposed neural network-based clustering approach serves as an effective representation of how machine learning algorithms can enhance decision-making structures. Keywords: inventory routing problem; integer programming-based heuristic; neural network; machine learning based decision-making
Pandemi sonrası Türkiye gün öncesi piyasasında doğal gaz fiyatlarının elektrik fiyatlandırması üzerindeki gecikmeli etkilerinin analizi
This study investigates the correlation between natural gas and electricity prices in Türkiye, which has experienced significant spikes in natural gas prices due to the pandemic and regional conflicts. The determination of reference prices for electricity and natural gas in Türkiye is facilitated by transparent markets operated by the Energy Exchange. Using data from the EXIST transparency platform, this research employs regression techniques to determine the temporal relationship between abrupt increases in natural gas prices and their impact on electricity prices. Prior studies have shown that elevated natural gas prices affect electricity prices with a delay, depending on the proportion of natural gas in a country's energy supply. This research applies a similar methodology to assess the time-lag effect of natural gas prices on electricity prices within the Turkish energy market post-COVID-19. Metrics such as mean absolute error, mean square error, mean absolute percent error, and mean percent error validate the findings. The results demonstrate a eightyfive-day time lag in the correlation between natural gas and electricity prices in Türkiye. This delay improves forecasting accuracy by up to 16.17% in terms of error reduction.
Teklif tabanli işletmeler arasi fiyatlandirma optimizasyonu
This paper focuses on quotation based pricing for a Business to Business (B2B) firm, using a real world case of a distributor engaged in aftermarket sales. The study assigns quotation groups into clusters, each associated with distinct logistic regression curves, to enhance overall performance and define the elasticities of various quotation groups. The objective of this study is to determine the optimal assignment of groups to clusters while simultaneously identifying the corresponding cluster wise regression coefficients. Coupled with the combinatorial nature of the task, finding an optimal solution is particularly challenging and computationally intensive. Additionally, we introduce a complex teacher model designed to improve the performance while using logistic regression model to enhance interpretability. To generate clusters, apart from using solvers, we propose heuristic algorithms such as the Recursive Partitioning Algorithm and the Iterative Refit Classifier Algorithm to address the problem. The performance of these approaches is evaluated using both real world and simulated datasets. Furthermore, we demonstrate the impacts of implementing quotation based pricing strategies using the real dataset.
Tahmin sonrası optimizasyon ve kısıt öğrenme: E-ticaret karar alma süreçleri için ampirik bir karşılaştırma
In today's rapidly evolving e-commerce landscape, effectively balancing price and commission strategies is vital for platforms aiming to maximize both merchant profitability and their own revenue. This MSc. thesis introduces an innovative approach by integrating advanced machine learning techniques with optimization frameworks, specifically comparing the traditional Predict-then-Optimize (PTO) method against the integrated Constraint Learning (CL) paradigm. Through rigorous simulation and empirical evaluation using real-world marketplace data, the proposed frameworks demonstrate clear benefits in strategic decision-making, efficiency, and robustness. Ultimately, this research paves the way for smarter, more dynamic pricing decisions in complex online retail ecosystems.
Fizik temelli sinir ağı tabanlı opsiyon fiyatlaması
Option pricing is an important problem in finance. The most common option pricing methodologies are Black-Scholes, CEV, Heston, and SABR, where all of them can be described by Partial Differential Equations. On the other hand, as a deep learning technique, Physics-Informed Neural Networks~(PINNs) have shown great potential in providing approximate solutions to Partial Differential Equations~(PDEs). In this study, we come up with Physics-Informed Neural Network-based option pricing methods under these 4 pricing models. We evaluate our method using both simulated and actual market data, and assess its performance against analytical and numerical reference models. The empirical results for index options such as RUT and SPX, as well as index constituent companies options, show that the proposed deep learning models have lower absolute errors than the classical Black–Scholes model solution. Our proposed solution also performs better than the recently proposed PINNsFormer for the option pricing task. Our results on both European and American options show the potential importance of Physics-Informed Neural Networks in solving option pricing, which can also be extended to other financial asset pricing problems. Our methods, analysis code, and datasets are available at: https://github.com/seferlab/PINN_for_Options
Sentetik katılımcılar: büyük dil modellerinin insan anket katılımcılarına alternatif olarak değerlendirilmesi
This thesis evaluates Large Language Models (LLMs) as "synthetic participants" in opinion surveys, addressing rising costs and biases in traditional data collection. Using 10,000 profiles from Wave 7 of the World Values Survey, ten LLM architectures were prompted to respond to seven trust-related questions. Performance was assessed via exact-match accuracy, mean absolute error (MAE), distributional divergence, and feature importance alignment, comparing LLM outputs to actual human participants and supervised machine learning models. Findings show that LLMs achieve a predictive accuracy that outperforms chance but lags behind specifically-trained supervised models. While errors were directionally accurate, divergence tests revealed a consistent "majority class fixation," where LLMs overrepresented dominant views and suppressed minority perspectives. Furthermore, while both machine learning models trained on human responses and those trained on LLM outputs rely on macro-institutional indicators, the first group is further shaped by cultural identity (ethnicity, language), whereas the second group relies on socio-economic status (income, social class). Consequently, LLMs function as "global average smoothers"—capturing broad trends but failing to replicate the heterogeneity required for high-fidelity social research. While promising for exploratory studies, LLMs remain an inappropriate substitute for human data in contexts requiring representational accuracy for minority or cross-cultural viewpoints.
Serviks hücrelerinin sınıflandırılmasında sürekli öğrenme stratejilerinin uygulanması
Medical imaging plays an essential role in clinical diagnosis and therapy planning. Machine learning is gaining popularity because it captures illness and therapy response characteristics. The continual improvement of image collecting technology and diagnostic methods, the diversity of scanners and developing imaging protocols, and field shifts restrict the usefulness of machine learning as predicted accuracy on new data degrades or models become out-of-date. An example for such shifts is pap smear imaging. This thesis presents applications of continual learning strategies for tackling models' outdating due to the domain shifts in pap smear images. Two popular convolutiunal neural network (CNN) architectures (InceptionV3 and ResNet50), and a custom 2-layer CNN are trained with three pap smear benchmark datasets (Sipakmed, Herlev, and CRIC). Several continual learning strategies are employed when training the models. Finally, the models for each strategy are evaluated, and the comparative results are presented
COVID-19'un tiroid hastalıkları üzerindeki etkisinin meta-analizinin yapılması
Millions of people were affected both by COVID-19 and thyroid diseases across the world but previous works are limited to identify the effects of COVID-19 on thyroid disorders. Thyroid holds a critical role to regulate numerous metabolic processes throughout the body, a few changes in it, may have affect on other systems on the human body. During the pandemic, thyroid disorders get increased and those disorders are usually interpreted as autoimmune mechanisms. Present meta-analyses are completed with limited parameters and just focus on a particular topic which prevents to act in a holistic approach. Our study focuses on the relation between thyroid function test measures, thyroid dysfunction, comorbidity and COVID-19 severity; also thyroid function test measures and mortality caused by COVID-19; thyroid function test measures and COVID-19 existence are observed. In total, 16 meta-analyses are completed and aim to improve the evidence level of the studies related to thyroid disorders and COVID-19 which will contribute to the medical sciences. 4260 studies were reviewed from PubMed and Science Direct search engines, and after applying the exclusion criteria, 13 studies with a sample group of 2829 individuals were included in the meta-analyses. PRISMA 2020 Checklist flows were tracked, for statistical calculation by using the RevMan and MedCalc software. All in all, thyroid dysfunction patients tend to undergo more severe COVID-19. The relation between thyroid disorders and COVID-19 was expressed with meta-analyses by using many parameters, aiming to increase the reliability level of the related studies.
Görüntü segmentasyonu için temel modellerin damıtılması
Recent advances in machine learning through model scaling achieves state-of-the-art results in various tasks. An example to this is foundation models, which are large models capable of generalizing well with zero-shot predictions. However, these models come with caveats of computational and memory costs. Knowledge distillation is a transfer learning technique where information from a bigger model can be distilled to a smaller architecture, saving memory and computational costs. In this thesis, we have investigated using foundation model as a teacher network in a knowledge distillation setting, compared to a large model fine-tuned on the same task.
Borsa İstanbul'da faaliyet gösteren teknoloji şirketlerinde lojistik regresyon modeli ile finansal başarısızlık tahminlemesi
Finansal başarısızlık ile karşı karşıya kalan her şirketin sonu iflas ile sonuçlanmaz. Bununla beraber, global düzeyde etki gösteren ekonomik krizin oldukça fazla oranda şirketin tasfiye edilmesine ve iflas etmesine sebep olduğu aşikardır. Finansal başarısızlık kavramı konusunda görüş birliği ve net bir tanımlama yoktur. Bu da, finansal başarısızlığın tahmin edilmesine dair araştırmaların iflaslarla ilişkilendirilmesine sebebiyet vermektedir. Finansal başarısızlık, uzun senelerdir şirketler, yatırımcılar ve kredi verenler bakımından en kritik tehdit olarak görülmektedir. Küreselleşmenin son yıllardaki etkisiyle, teknolojik ilerlemelerin ve finansal koşulların da yansımasıyla ulusal ve uluslararası sektörlerin hızla farklılaşmakta ve ilerlemekte olduğu görülmektedir. Bu farklılık ve ilerleme, daha çok şirketler bakımından, finansal başarısızlıkları tahmin etmeyi ön plana çıkarmaktadır. Bu çalışmada finansal başarısızlık kavramı, finansal başarısızlık çeşitleri, finansal başarısızlığın önlenmesi konusunda alınabilecek tedbirler, finansal başarısızlığın tahmininde kullanılan yöntemler, finansal teknoloji, finansal teknoloji süreci, finansal teknoloji oyuncuları ve Borsa İstanbul kavramları kavramsal olarak anlatılmıştır. Çalışmada finansal veriler ile şirketlerin başarısızlıklarının lojistik regresyon yöntemi ile tahmin edilmesi amaçlanmıştır. Borsa İstanbul'da işlem gören 27 adet teknoloji şirketine ait veriler toplanarak rasyolar oluşturulmuştur. 2021 verileri ve son üç yıldaki finansal veriler ile başarısız şirketler belirlenmiştir. Şirketlere ait veriler öncelikle3 aylık dönemlere göre oluşturulmuş fakat bazı şirketlerin 2018 ve 2019 yıllarına ait sadece yıllık verilerine ulaşılabilmesi sonucunda yıllık veriler kullanılmasına karar verilmiştir. 2018, 2019 ve 2020 yıllarına ait belirlenen rasyolar bağımsız değişken olarak kullanılmıştır. Eksik verilerin bulunduğu rasyolar çıkarıldığında 60 farklı rasyo ile analiz yapılmıştır. Şirketlerin finansal başarısızlığının önceden tahmin edilmesinde 3 yıl önceden tahmin etmede Duran Varlıklar/Maddi Özkaynak, Net Borç/FAVÖK(Yıllık) ve Toplam Borç/Özsermaye rasyoları, 2 yıl önceden tahmin etmede FAVÖK/Büyüme (Yıllık) ve Kısa Vade Borç/Toplam Borç rasyoları ve 1 yıl önceden tahmin etmede Özsermaye Karlılığı/ROE ve Kısa Vade Borç/Büyüme rasyolarının başarılı olduğu sonucu ortaya çıkmıştır.
Türkiye otel verileri üzerinde kümeleme ve öneri sistemi
As in most sectors, the development of an intelligent recommendation system in tourism becomes an important issue. Tourism agencies are putting maximum effort into suggesting the best and most valuable hotels for their customers. With the help of B2B relations between agencies and hotels, tourism agencies hold large feature datasets about hotels. Summarizing or interpretation of huge amount of data, requires the implementation of data analysis methodologies. Also, the tourism data is unique in terms of geography and culture. Thus, every new dataset requires a dedicated analytical process. Furthermore, raw data is in the form of a sparse binary matrix of hotel features, it poses a technical challenge for any analytical process. This thesis presents a comparison of different clustering and dimension reduction methodologies for real-world hotel data of this nature. The dataset represents 61% of the hotels in Turkey. Hotel clustering is the first step to acquire the hotel recommendations. To generate matching recommendations for customers, multiple level system is designed. Another layer of this cascade structure is clustering of the users according to their previous hotels moreover clusters of hotels. Our approach has resulted with linkage of two type of collaborative filtering which called as hybrid recommendation system. Thereby the suggested system provides personalized hotel recommendations based on the hotel's amenities and visitor patterns. The first challenge in this work is the nature of the raw data set. The hotel features that we work with are all binary variables. While there are plenty of metrics, algorithms, and techniques dedicated to discovering knowledge from numerical variables, the methods for processing binary ones are limited. Thus, one of our contributions is to propose an experimental methodology for discovering the best clusters for the hotels, which are explained with binary features. In this methodology, we transform the sparse binary data set into numerical ones by using different dimension reduction techniques. Then, well-known clustering algorithms are applied and evaluated by various success criteria metrics. The most succeeded clustering algorithm has been decided as OPTICS. Due to the nature of the algorithm, noise labeled hotels has been created by the results of the algorithm. Cosine similarity between hotel features is calculated and the result has been used for elimination of the noise labeled hotels. Since this is the first step in the process, rather than focusing on the interpretation of preliminary clusters, we have focused on solving analytical problems such as determining the number of the best clusters, identifying the most distinguished clusters, and simply selecting the best algorithm for hotel clustering. During the recommendation engine design, we used collaborative filtering method which consist of item and user features. Thereby the engine can be considered as a hybrid system. According to designed similarity matrix between users, the target user (customer) receives a bunch of last visited hotels. Similarity between users has been decided based on their clusters of already visited hotels. Thereby every user has at least one similar user, and these users have at least one order in our analysis table. While creating recommendation arrays, the results checked with sales test table which consist only latest orders. If the test value for target customer matches with one of the members of recommendation array, the recommendation considered as success. Even though the binary labeled success criteria (success or fail) which is very loose method, the recommendation engine has not achieved the expected success ratio. At last, data preparation steps, used models and recommendation system results has been discussed and the reasons of the success rates are explained in this thesis.
COVİD-19 sürecinde hareketlilikte eşitsizlik: Küresel ve yerel analiz
This thesis analyzes mobility patterns during the Covid-19 pandemic from a global and local perspective. The global framework includes 37 European countries and the local framework comprises 81 Turkish cities. The study follows the daily mobility trajectories of people from February 2020 to January 2022. The analyzes are conducted to understand the economic opportunities available in countries -at a macro scale- that facilitate or hinder the "proper" mobility behavior of individuals while focusing on the captive commuters, i.e., the share of the population who need to commute to the work despite the risk of infection and governmental policies. The results indicate that the workforce in regions with higher GDP per capita, education level, and life expectancy at birth was able to reduce their workplace mobility higher than commuters in areas with low income, education level, and life expectancy at birth. Therefore, unprivileged populations were exposed to higher health risks against rapid Covid-19 transmission in Europe and Turkish cities.
Geliştirilmiş RFM modeli ile müşteri segmentasyonu: Bir halı ve kilim üretici firmasında uygulama
Data science has gained enormous importance by contributing to the in-depth understanding and interpretation of information. Especially companies consult on data analysis to make strategic decisions in the competitive market. Much more important than the decisions taken is a determination of the customer or customer groups to which these decisions will be adapted. For that reason, customer segmentation by identifying similarities and differences between customers becomes crucial. In recent times, the RFM model is preferred mostly for customer segmentation. The RFM model is based on the customer's last purchase date, how often they purchase, and how much money contributes to the company. It is an easy model to understand and interpret results in a clear way. Many researchers prefer to apply the RFM method by adding extra variables to the analysis. Thus, customers are evaluated from a broader perspective. This study aims to present a developed RFM model by adding extra variables which are Loyalty, Dependence, and Expectation which are determined by a broad literature review and as a result of a survey relating to 106 dealers. There are some studies that create a segmentation model by using loyalty and the RFM model. However, this study developed a new model by including the dependence and expectation variables, which are not been used previously with the RFM model, besides loyalty. In the study, dealers are analyzed by the K-means clustering method and the optimum number of clusters is indicated as six. Each cluster has its specific customer behavior and this study guides the company to constitute marketing strategies regarding customers' specifications.
Üniversitelerde ar-ge harcamalarının araştırma performansına etkisi
The most significant factor showing that countries are economically developed is the level of technology they produce. Technology, on the other hand, is the application of knowledge produced theoretically through scientific research. In this context, there is a clear connection between the level of economic development and welfare and the point reached in science. The resources that countries allocate to science and research activities are increasing every year, and the effective use of these resources plays a critical role in increasing the welfare level of these countries. This study reveals the effect of the university investment budget, which constitutes an essential part of the financial resources allocated to research and development activities in Turkey, on the research performance of universities. Multiple regression analysis was used to achieve the aim outlined in this thesis. The number of publications per lecturer published in AHCI, SSCI, SCI, and SCI-Expanded indexed journals and the number of Web of Science publications citations were used as dependent variables to represent research performance. The rate of investment budget spent on R&D was used as an independent variable to represent R&D expenditures. In addition, the number of lecturers and the number of YÖK 100/2000 Project doctoral scholars were included in the models as control variables. It was concluded that R&D expenditures affect research performance positively.
COVID-19 döneminde Türkiye ve İsveç'te uygulanan politikaların fırsat eşitliği ve eğitim çıktıları üzerine etkisi
This thesis delves into the impact of COVID-19 policies implemented in Türkiye and Sweden on equal opportunity within higher education and explores the consequential higher education outcomes. The research employs a mixed-methods approach, incorporating both qualitative and quantitative methodologies. Through an examination of socio economic structure, education system, the study elucidates the distinctive approaches taken by Türkiye and Sweden. The findings contribute significantly to a comprehensive understanding of global education policy responses during crises, emphasizing the pivotal role of ensuring equal opportunity. By scrutinizing the specific measures undertaken by both countries, this study not only informs the on education policy during extraordinary times but also provides valuable insights for policymakers, educators, and stakeholders seeking to enhance equal opportunity and foster positive outcomes in higher education.
Istatistiksel ön puanlama bileşeni ile gruplama puanlama modellemesi (GSM) yaklaşımın geliştirilmesi: Yüksek boyutlu transkriptomik veri analizi için bir vaka çalışması
Rapid advancements in transcriptomic technologies have significantly increased the volume of data available for analysis, which presents challenges in terms of efficiency and computational demand. This thesis introduces a Pre-Scoring component to the Grouping-Scoring-Modeling (G-S-M) framework to address inefficiencies caused by the excessive number of gene groups generated by traditional GSM. By selectively prioritizing gene groups based on their statistical significance, this innovation aims to reduce the computational demands associated with scoring these groups using machine learning models, thereby streamlining the analysis process. Assessed across nine diverse Gene Expression datasets, the Pre-Scoring G-S-M framework not only maintained accuracy comparable to the traditional approach but did so with significantly fewer genes. This refinement conserves resources while maintaining the robustness and reliability of the data analysis, crucial for advancing research in personalized medicine and therapeutic strategies. The findings suggest that the modified G-S-M framework serves as a valuable tool in bioinformatics, offering a more efficient approach to handling large-scale genomic datasets. Future work will focus on adapting this enhanced framework to incorporate diverse types of omics knowledge, such as proteomics and metabolomics, further optimizing its performance to broaden its applicability in both clinical and research settings
Ulusal gönüllü değerlendirme raporları üzerinden sürdürülebilir kalkındırma hedeflerine ulaşma analizi
The 2030 Agenda for Sustainable Development sets 17 Sustainable Development Goals (SDGs) to address critical global challenges such as poverty, inequality and climate change. To monitor progress towards these goals, countries submit Voluntary National Reviews (VNRs) that qualitatively and quantitatively assess their SDG implementation. This study comprehensively analyses the VNR reports submitted in 2023 and 2024 and examines the extent to which these reports reflect progress towards the SDGs. Using a multi-layered text classification model supported by advanced machine learning algorithms, the alignment of VNR reports with specific SDGs is categorized and assessed. The study provides data-driven recommendations on regional differences, performance by goal, and the development of reporting frameworks. The findings show that SDG 5 (Gender Equality) and SDG 13 (Climate Action) are prioritized in VNR reports in many countries, while SDG17 (Partnerships for the Goals) is underrepresented in VNR reports. In addition , positive and negative relationships between the goals have been addressed in VNR reports. With these results, recommendations for more effective policy development, international cooperation and data-driven decision-making processes are presented for decision-makers.
Türkiye'de dindarlığın milliyetçilik ve küreselleşme tutumları üzerindeki etkilerinin araştırılması: Dünya değerler araştırması verileri kullanılarak yapılan bir çalışma
This study examines the complex relationships across religiosity, nationalist and globalist attitudes in Türkiye. The study provided a unique conceptual foundation for Türkiye by examining whether religiosity supports nationalist and globalist attitudes. This study analyzes how religion affects nationalism and globalization through the variables of attendance to religious services, national pride, and support for the European Union (EU) and the United Nations (UN). Using the Turkish leg of World Values Survey (WVS) data, and a series of generalized ordered logistic (gologit) regression models, this study aims to provide empirical findings revealing relationships between these variables. Understanding the associations between these variables provides crucial insights into the question what determines participants' sense of national identity and belonging to global institutions. The findings presented that religiosity interacts positively with national pride, and negatively with sonfidence with the EU.
Makine öğrenmesi teknikleri kullanarak moda e-ticaret sektöründe müşteri segmentasyonu
In today's world where technology is developing very rapidly, internet usage is also increasing proportionally. This change has revealed that brands attach importance to the sector. The significance of e-commerce is to the advantage of brands because there have been decreases in some fixed expenses of companies. With the increase in online shopping, personal analyses of customers can also be made by customer relationship management (CRM). It is necessary to divide customers into segments for customer-oriented marketing. Customer segmentation is a widely used form of analysis. There is an increasing demand to develop a deep awareness of individual customer needs and desires. Segmentation, a commonly utilized method for achieving this understanding, has undergone continuous refinement in recent years. This study targets to present a detailed analysis of various segmentation approaches and their evolution. In this study, RFM (Recency, Frequency, Monetary) analysis was used for segmenting the customers. Customers were divided into segments by scoring them on the last shopping time, shopping frequency and total spending. Four customer groups were created with K-means and the values of each segment were analyzed. Churn rate analysis determined customers who did not shop for 90 days as lost. Churn estimation was performed with the LightGBM model using the machine learning technique. In addition, the Predictive CLV (customer lifetime value) model was developed using the 1 machine learning technique Ridge Regression. The accuracy rate was increased and low, medium and high CLV segments were created. As a result; RFM , K-means and CLV estimation were used to optimize customer relationships and increase revenues. E-commerce data of a private brand was analyzed using machine learning techniques. Nowadays, there is an increase in computing power and rapid developments in machine learning/artificial intelligence algorithms. This has recently enabled the application of more advanced techniques.
Reklam değerinin yapay zeka tarafından oluşturulan reklamlara yönelik tutum ve tüketici satın alma niyeti üzerindeki etkisi: Planlı davranış teorisi perspektifi
Artificial intelligence is one of vital technological tools increasingly popular andused by companies in marketing applications such as advertising. This thesis aims toinvestigate relationship between AI-generated ads and Ad value, attitudes, and purchaseintentions, and moderating roles of variables such as gender and self-efficacy. Thetheoretical framework has been designed in line with Theory of Planned Behavior andAdvertising Value Model, which provides a comprehensive perspective on howindividuals evaluate and respond to AI-supported promotional content. Data was collectedthrough an online survey from participants aged 18-64 between August and September2025, yielding 361 responses. IBM SPSS Statistics 25 and Python were used for dataanalysis. Reliability analyses, Factor Analysis (EFA, CFA) were performed to verify theadequacy and robustness of the measurement tools, and hypotheses were analyzed withStructural Equation Modeling (SEM). According to results perceived ad value positivelyaffects both attitudes toward AI-generated ads and purchase intentions. Attitudes towardAI-generated ads notably affect purchase intentions. Attitudes have a mediating role inthe relationship between purchase intentions and perceived advertising value. Whilegender had no moderating impact on perceived ad value and attitude relations, itmoderated ad value and purchase intention relations. Its effect on the relationship betweenattitude and purchase intention is not statistically significant. Self-efficacy significantlynegatively moderates the relationship between attitudes and Ad Value. While it positivelystrengthens the effect of advertising value on purchase intentions, it negatively affects the effect of attitude on purchase intention. Theoretically, this research advances knowledgein AI-enabled marketing communications by integrating two key moderating individualcharacteristics into the relationship between value perceptions, attitudinal responses, andbehavioral outcomes. Also, it provides managerial benefits by highlighting the importanceof personalizing AI ads based on consumer behavior.
Kredi hacmi ve takipteki kredilerin kısa vadeli ve uzun vadeli analizi: Türk bankacılık sektörü perspektifi
This thesis investigates the dynamic relationship between loan volume and NPLs in the Turkish banking sector, focusing on both short-term and long-term perspectives. Using weekly data from January 2014 to May 2024 and employing the DCC model, this study explores how credit growth and economic shocks influence NPLs. The findings reveal a significant correlation between loan volumes and NPLs in the long term. On the other hand, the thesis also shows notable fluctuations during financial instability, such as the 2018 currency crisis and the COVID-19 pandemic. The cross-correlation analysis further highlights the effects of loan growth on NPLs following the importance of proactive risk management. The research provides actionable insights for policymakers, emphasizing the need for balanced credit growth, enhanced regulatory frameworks, and dynamic risk assessment tools. These findings contribute to the literature by offering a nuanced understanding of the relationship between credit dynamics and financial stability in an emerging market context. The study also aligns with Sustainable Development Goals (SDGs) 8 10, and 17, supporting inclusive economic growth and reduced financial inequalities and partnership through improved banking practices. Keywords: Non-Performing Loans, Loan Volume, Banking Sector
Gücün terapi odasında sürdürülmesi: Sosyal baskınlık yönelimi, medikalizasyon ve aşırı patolojikleştirmenin kavramsal modeli
This thesis examines how therapists' orientation towards social hierarchies and their medicalization beliefs shape the risk of overpathologizing in psychotherapy. Drawing on Social Dominance Orientation (SDO), it develops a conceptual model in which SDO influences clinical judgement through medicalizations, understood as an ideological framework that legitimizes expert authority. Study 1 analyzed 1,675 naturally occurring client complaints from a public platform using large language model-based text classification. The complaints were coded for overpathologizing and grouped into the themes of overprescribing, overdiagnosis, pessimistic prognosis, and coercive treatment, providing ecological evidence that overpathologizing is a repeated problem in professional practice. Study 2 employed an online vignette experiment designed to activate SDO would increase medicalization beliefs and, in turn, overpathologizing tendencies. The manipulation produced no group differences, but medicalization beliefs were positively associated with overpathologizing, and post hoc analyses suggested an indirect pathway from SDO to overpathologizing via medicalization. Together, these findings emphasize overpathologizing as both an empirical reality and a theoretically explicable bias rooted in hierarchy-maintaining ideologies. Positioning clinical judgement within broader dynamics of power and legitimizing discourse, the thesis advances a framework for understanding how institutional authority can subtly translate into client harm, underscoring the ethical imperative for power-aware psychotherapy.
Türkiye'nin yenilenebilir enerji potansiyelinin ve çevre protokolü uyumluluğunun değerlendirilmesine yönelik bir makine öğrenimi
The issue of global warming has been identified as one of the most critical challenges of the 21st century, with the consumption of fossil fuels being identified as a major contributor to greenhouse gas emissions. In response to these challenges, countries worldwide are expediting their transition towards renewable energy sources to meet international climate commitments, such as the Paris Agreement, and to achieve long-term sustainability goals. Türkiye has set itself the target of reaching net-zero emissions by 2053, a goal which is in alignment with both its national energy strategy and its international obligations. Nevertheless, due to geographical, economic and technological constraints, the transition from fossils to renewable energy sources is challenging. The present study aims to assess the capacity and efficiency of renewable energy in Türkiye with environmental protocols and future electricity demand projections. Electricity generation-transmission data and national energy plans are used to identify future electricity generation and capacity trends. In the context of this study, a range of machine learning models are executed across diverse scenarios, yielding a series of outcomes. Consequently, the repercussions of regulatory measures and financial investments were examined, and prospective inferences were derived. The findings emphasize the significance of scenario-based modelling in the formulation of sustainable energy policies and the guidance of investment decisions within the context of climate change mitigation.
Yatırımcı tipleri ve riskten kaçınma: Finansal karar vermede veri bilimi yaklaşımı
This study examines the manifestation of behavioral biases in investor decisionmaking through a quantitative data science approach using the FAR-Trans dataset. We developed empirical measures for three key behavioral finance constructs (loss aversion, herding behavior, and overconfidence) derived from established theoretical frameworks and applied them to classify investors into distinct behavioral groups. A decision tree-based multi-class classification model was employed to predict behavioral bias categories, demonstrating exceptional predictive performance with 96% accuracy and macro-average ROC-AUC scores of 0.99 validated through crossvalidation analysis. Statistical significance tests and visual analyses confirm that each behavioral group exhibits clearly distinguishable patterns in their respective dominant metrics, with loss-averse investors displaying higher tendencies to maintain losing positions, herding investors showing increased propensity to follow popular assets, and overconfident investors engaging in more frequent trading activities. These findings provide robust empirical evidence for the effective identification and classification of behavioral biases from transactional financial data and demonstrate the practical utility of machine learning techniques for understanding psychological factors in financial markets.