Theses supervised by Adnan Acan
13 theses · Eastern Mediterranean University
Data Modeling with Type I and Type II Fuzzy Sets
The fuzzy classifier is an algorithm that assigns a class label to an object, based on the object description. It is also said that the classifier predicts the class label. The object description comes in the form of a vector containing values of the features (attributes) deemed to be relevant for the classification task. Typically, the classifier learns to predict class labels using a training algorithm and a training data set. When a training data set is not available, a classifier can be designed from prior knowledge and expertise. Once trained, the classifier is ready for operation on unseen objects.In this thesis, type-1fuzzy classifier, and the type-2 fuzzy classifier are used for the machine learning datasets classification. The Wisconsin breast cancer dataset, Iris Dataset, and Tic-Tac-Toe datasets are classified. Type-2 fuzzy classifiers are able to perform better than type-1 fuzzy classifiers which have additional design parameters. Therefore, type-2 fuzzy classifiers are more attractive than the type-1 classifiers. The essential benefits the type-2 fuzzy logic classifiers are their ability to handle more vagueness. Keywords: Classifier, Type-1fuzzy classifier, Type-2 fuzzy classifier, Machine learning dataset and Uncertainty.
Evaluation of Surrogate Assisted Differential Evolution Algorithm for Single-Objective Numerical Optimization
Hard optimization problems are solved successfully using nature inspired metaheuristics. However, in many cases of practical optimization problems, also called black-box problems, the evaluation of the objective function is main cause of high demand of computational resources. In the solution of these problems, objective function landscape is modeled mathematically, called a surrogate model which consist of replacing the objective function by an equivalent mathematical model, to reduce the computational evaluation time of the fitness function. The differential evolution (DE) algorithm is implemented with 4 strategies called rand/1, rand/2, best/2 and rand to best/1 to optimize the benchmark functions listed CEC2017 competition with dimensions D=10 and D=30. CEC2017 benchmark set is composed of 30 different functions with different degree of complexities. Locations of optimal solutions for these functions is supposed to be unknown and that’s why they are called black box functions. A surrogate model called the quadratic response surface model (QRSM) is used with Latin hyper square sampling strategy to replace objective function evaluations of benchmark functions. QRMS is used with DE for the solution of CEC2017 benchmark problems for the purpose of evaluating the performance of the surrogate assisted DE algorithm in terms of solution quality and runtime complexity. Experimental results obtained from the 4 different DE and DE+QRSM strategies illustrated that the rand/1 DE strategy was generally the best strategy in speed and accuracy for both dimensions D=10 and D=30. Also, the results generated by DE and DE+QRSM are compared with each other. As illustrated in tables of experimental evaluations, DE is found more accurate in majority of benchmark functions but it is slower generally. Also, a comparative study is done with other published algorithms such as L-SHADE, JSO, DISH, L-SHADE-LBR, JSO-LBR and DISH-LBR. Results obtained by these competitors are compared to only the best DE strategy, which is rand/1, employed within DE and DE+QRSM. The rand/1 strategy implemented within DE function was quit robust and performed better than other algorithms in many cases for D=10, but when implemented within DE+QRSM it becomes the worst one. For D=30 the rand/1 strategy loosed of its performance and was classified before the last position. Its rank is around of 80% when implemented within DE but it stays in last position with DE+QRSM algorithm.
Adaptive Differential Evolution Algorithm for Single and Multi-Objective Numerical Optimization
“DE/current-to-pbest” is a new and increasingly common mutation strategy that involves an additional external archive and adaptively updates the control. This thesis introduces a novel algorithm known as JADE. The “DE/current-to-pbest” is a simplification of the typical “DE/current-to-best,” while historical data is used by the additional archive operation to provide information on progress direction. Both convergence performance and the diversity of the population are enhanced by the two operations. The control parameters are automatically updated to the appropriate values through parameter adaptation, which avoids relying on outdated information regarding the relationship between the characteristics of the optimization problems and the parameter settings. This thesis work introduces a JADE Algorithm and examines its feasibility based on the results of CEC'17 expensive benchmark problems for single objective optimization problems and for Multi-objective optimization. The methods used in our studies are compared to different well-knows methods proposed in the related literature was conducted. The final ranking of all test problems indicate that JADE was always among the top best algorithms that were used for the same purpose.
Face Recognition Using Random Forest Classifiers Based on PCA, LDA and LBP Features
Face is the main part of human beings to distinguish from one another. Face recognition system mainly takes an image as an input and compares this image with a number of images stored in the database to identify whether the input image is in the database or not. Also, face recognition is the process of identification and verification of individuals by their facial images. In this thesis, well-known databases such as FERET and JAFFE databases are used for experimental evaluations. Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA) and Local Binary Patterns (LBP) are used for extracting facial features of individuals from the region of interests. Decision Tree (DT) and Random Forest (RF) are used as classify the faces based on extracted features. The Manhattan Distance measure is used to compare the difference between test and training images for face recognition. Based on the experimental evaluations, the achieved recognition rates are very close to those published articles in the literature. Keywords: Local Binary Patterns (LBP), Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), Random Forest (RF), Decision Tree (DT), feature extraction, classification.
A Distributed Multi Event Solution for Recommender Systems Using Hadoop
Big data is a phenomenon that takes central stage in industry and academia arising from the advent of online services and mobile applications. Improving the efficiency of data processing and analysis has become a challenging issue. While a number of methods from different communities have been proposed for solving the “Big Data” problems, we worked with multi-event Intelligent Systems that offer efficient mechanisms, which significantly reduce the costs of processing large volume of data and improve data processing quality. Social networks could benefit from recommender systems in order to optimize the queries and ads they display for each special user. Among different approaches to analyze user data and making recommendations, we employed Collaborative Filtering with Cosine Similarity criterion for item-based similarity recognitions. In the implemented method, a Holonic multi-event system (HMES) is designed to process a portion of Amazon database in a distributed manner. The use of Hadoop and map-reduce technology is aimed to make more accurate and faster predictions and recommendations. Different evaluation standards such as Perfect Hit (PHIT), and Mean Percentage Rank (MPR) are used to examine and compare the proposed method with other conventional methods. The results obtained in this thesis are satisfactory compared to the results of the evaluation given in the literature. Keywords: recommender system, hadoop, multi event, artificial intelligence, big data, holonic
Medical Image Enhancement through Intuitionistic Fuzzy Sets
A contrast enhancement of medical, color and Grayscale, images via intuitionistic fuzzy sets on different types of entropy – based methods have been studied. Fuzzy set concept counts vagueness in the formula of the membership functions. Intuitionistic fuzzy sets count fuzziness in the membership and non-membership functions. Various entropy – based methods are applied as enhancement operators, and the enhanced image is the one that is interpreted based on the used intuitionistic fuzzy membership function. As medical images include too much ambiguity, this study demonstrated that the intuitionistic fuzzy sets are shown to be useful tools implemented for medical image enhancement. To determine the efficiency of the studied methods, experimental results associated with the handled entropy methods are presented in thesis. Experiments on several image libraries indicate that the spatial entropy method among several applications performs better than it is alternatives. Keywords: Intuitionistic Fuzzy Set; Fuzzy Entropy, Spatial Entropy, Contrast Enhancement, Image Quality Measurement.
Deep Learning Based Processing of EEG Signals for Detection and Recognition of Parkinson Disease
The aim of this study is to provide early detection of Parkinson's disease by processing EEG signals through two dimensional colored image transforms. Parkinson's disease is a neurological disease that usually occurs in old ages and occurs with a decrease in dopamine levels in the brain. There is no known treatment for Parkinson's disease. Early detection and early treatment in Parkinson's disease is very important to slow the progression of the disease. EEG data were obtained from the UC San Diego Resting State EEG Database from Patients with Parkinson's disease. EEG signals were converted to GASF images by going through various preprocessing steps. AlexNet deep learning model was used to train and test the obtained 2D colored image data. AlexNet is a Convolutional Neural Network model consisting of 8 layers. In the literature review, 16 channels used in various studies were selected. Amoung these Fp1, F7 and F3 channels are the ones with highest reported succes results. The same channels are also considered with in the scope of address in this thesis work. GASF images of selected Fp1, F7 and F3 channels were used to train the AlexNet CNN model over 100 epochs. The developed model achieved promising performance with 97.72% accuracy, 97.76% sensitivity and 97.68% specificity. In addition, the AlexNet CNN model was trained and tested over 100 epochs with 4-fold Cross Validation. As a result of this study, the developed model achieved the highest results with 97.73% accuracy, 97.94% sensitivity and 97.53% specificity.
Multiagent Coordination Using Probability Collectives
This thesis motivates and describes the use of probability collectives (PC) with a multiagent coordination system to solve different problems. The main challenge was to enable the agents to work in a coordinated way, optimizing the local utilities and contributing the maximum or minimum towards optimisation of a global objective. The approach was validated solving numerical benchmark problems such as sphere function in which the coupled variables are seen as autonomous agents working collectively to achieve the optimum solution. Moreover, PC algorithm solved successfully repeated games such as prisoner‟s dilemma, stag hunt, the battle of sexes game and choose sides. In all experimental trials, the optimum results were obtained at a reasonable computational cost. Keywords: Probability Collectives, Collective intelligence, Multiagent systems, Game theory.
Bee Colony Optimization for Single and Multi-Objective Numerical Optimization
One common feature of natural systems is the ability for the dynamic interaction between the most basic individual organisms to produce systems capable of performing complex tasks. This thesis introduces a novel population based search algorithm known as Bees Algorithm (BA), that simulates the manner in which swarms of honey bees forage for food. This algorithm involves a collection of a neighborhood and stochastic search and is used in both functional and combinatorial optimization. After describing the algorithm in detail, this thesis attempts to elucidate the robustness and efficiency of the algorithm based on the outcomes for a library of complex numerical optimization problems. The Artificial Bee Colony (ABC) is a swarm based on meta-heuristic algorithm used to optimize numerical optimization problems and provide accurate solutions. The use of the term ‗meta-heuristic‘ here refers to the capacity of the algorithm to provide optimal solutions even in cases on imperfect or incomplete information. Bee colonies scour many sources of food to determine the best source based on a number of parameters, such as time, the amount and quality of nectar, etc. In a similar manner, models that use the ABC algorithm are composed of three components: Unemployed bees, Employed bees, and Food sources (Fitness). The employed bees are responsible for finding affluent sources of food close to the hive. In the algorithm, artificial forager bees acting as environmental agents search for rich food sources. The process of applying the algorithm begins with transforming the given optimization problem into one of examining the best parameter vectors, from a population of vectors, to minimize the objective function. Starting with population of preliminary solution vectors, potential solutions are enhanced using certain strategies. This thesis work introduces a Bee Colony Optimization Algorithm and examines its feasibility based on the results of CEC'05 and CEC'17 expensive benchmark problems for single objective optimization problems , and used CEC'09 and CEC'18 expensive benchmark problem for Multi-objective optimization. The methods used in our studies are compared to different well-knows methods proposed in the related literature was conducted. The final ranking of all test problems indicate that BCO was always among the top best algorithms that were used for the same purpose. Keywords: Multi-agent systems, Meta-heuristic algorithms, Multi-objective optimization, Swarm intelligence, Pareto optimality
Cooperative Multi-agent Systems for Single and Multi-objective Optimization
Solving combinatorial and real-parameter optimization problems is an important challenge in all engineering applications. Researchers have been extensively solving these problems using evolutionary computations. In this thesis, three new multi-agent architectures are designed and utilized in order to solve combinatorial and realparameter optimization problems. First architecture introduces a novel learning-based multi-agent system (LBMAS) for solving combinatorial optimization problems in which all agents cooperate by acting on a common population and a two-stage archive containing promising fitness-based and positional-based solutions found so far. Metaheuristics as agents perform their own method individually and afterwards share their outcomes with others. In this system, solutions are modified by all running metaheuristics and the system learns gradually how promising metaheuristics are, in order to apply them based on their effectiveness. In the second architecture, a novel multi-agent and agent interaction mechanism for the solution of single objective type real-parameter optimization problems is proposed. The proposed multi-agent system includes several metaheuristics as problem solving agents that act on a common population containing the frontiers of search process and a common archive keeping the promising solutions extracted so far. Each session of the proposed architecture includes two phases: a tournament among all agents to determine the currently best performing agent and a search procedure conducted by the winner. The proposed multi-agent system is experimentally evaluated using the well-known CEC2005 benchmark problems set. The third architecture presents a creative multi-agent and dynamic multi-deme architecture based on a novel collaboration mechanism for the solution of multiobjective real-parameter optimization problems. The proposed architecture comprises a number of multi-objective metaheuristic agents that act on subsets of a population based in a cyclic assignment order. This multi-agent architecture works iteratively in sessions including two consecutive phases: in the first phase, a population of solutions is divided into subpopulations based on the dominance ranks of its elements. In the second phase, each multi-objective metaheuristic is assigned to work on a subpopulation based on a cyclic or round-robin order. The proposed multiagent system is experimentally evaluated using the well-known CEC2009 multiobjective optimization benchmark problems set. Analysis of the experimental results showed that the proposed architectures achieve better performance compared to majority of their state-of-the-art competitors in almost all problem instances. Keywords: Multi-agent systems, Metaheuristics, Combinatorial Optimization, Multiprocessor Scheduling, Agent Interactions, Multi-objective Optimization, Pareto Optimality
Stock Market Prediction Using Analytic Hierarchy Process and Support Vector Machine
Prediction of the stock market behavior has been a research topic for decades. Because it is a challenging subject both in terms of the choice of the prediction model and in terms of constructing the set of features that model will use for forecasting. In this thesis, a novel feature ranking and feature selection approach incorporation with weighted kernel least squares support vector machines (LS-SVMs) were used. We introduce the analytic hierarchy process (AHP) into the stock market and then evaluate criteria which provide the prediction model with relevant knowledge of the underlying processes of the studied stock market. The feature weights obtained by the AHP method are applied for feature ranking and selection and used with the LS-SVMs through a weighted kernel. The experimental results specify that the new model outperforms the benchmark models. Furthermore, the set of feature weights obtained by the new approach can also independently be incorporated into other kernel-based learners. Keywords: stock market prediction, analytic hierarchy process, support vector machine, least squares support vector machines, weighted kernel.
Fusion of Multi-stage CNN Features for ECG Classification
ABSTRACT: Detecting and classifying cardiac arrhythmias is critical to the diagnosis of patients with cardiac abnormalities. Identification and classification of abnormalities are time consuming because it often requires analysing each heartbeat of the ECG recording. Moreover, computerized ECG classification can also be very useful in shortening hospital waiting lists and saving the life by discovering heart diseases at early stages. Therefore, automatic classification of the arrhythmias using machine-learning technologies can bring various benefits. In this thesis, novel and high-performance approaches based on deep learning techniques are proposed for the automatic classification of electrocardiogram (ECG) signals. In this research work, two fully automatic systems have been presented which are shown to have high efficiency and low computational cost. In one of the proposed systems, a novel decision-level fusion of features is presented by three different approaches; the first one uses normalized feature-level fusion of handcrafted global statistical and local temporal features by uniting these features into one set, the second one uses the morphological feature subset, and the third one combines features extracted from multiple layers of a Convolutional Neural Network (CNN) through using a score-level based refinement procedure. The second proposed system utilized a new architecture of deep neural networks, Directed Acyclic Graph Convolutional Neural Networks (DAG-CNNs). DAG-CNNs fuse the feature extraction and classification stages of the ECG classification into a single automated learning procedure and utilize the multi-scale features and perform the score-level fusion of multiple classifiers automatically. The results over the MIT-BIH arrhythmia benchmark database exhibited that the proposed systems achieve superior classification accuracy compared to all of the state-of-the-art ECG classification methods. Keywords: electrocardiogram, convolutional neural networks, directed acyclic graph CNN, morphological feature, statistical feature, temporal features, multi-stage CNN-based features, feature-level fusion, score-level fusion, decision-level fusion
Teaching-Learning Based Algorithm for Numerical Dynamic Multi-Objective Optimization Problems
Teaching–Learning-Based Optimization (TLBO) algorithm has become an alternative optimization method in a great number of applications in different fields of engineering and science since it has been introduced in 2011. Teaching-learning-based optimization (TLBO) is a population-based metaheuristic examination algorithm stimulated by the teaching and learning procedure in a classroom environment. TLBO with its comparatively reasonable performances outperforms some of the well-known metaheuristics concerning constrained benchmark tasks, controlled mechanical schemes, and nonstop non-linear numerical optimization problems. In the TLBO algorithm’s variants, all the learners have an equal chance of receiving information from the teacher and from each other as learners by interacting with each other in the class. The Experimental results of the TLBO method are tested on the set of CEC2018 dynamic multi-objective optimization benchmark problems and the computed results show that TLBO offers promising outcomes with diverse dynamic features and changing environments compared with other algorithms. It works well with producing a good class of population when alterations happen for pursuing the influential Paretooptimal set efficiently for refining population conjunction and multiplicity. TLBO extracted improved or equal quality solutions compared to other evolutionary algorithms. It is a promising alternative for the solution of difficult dynamic multiobjective optimization problems.