Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı Tez Koleksiyonu

Permanent URI for this collectionhttps://hdl.handle.net/20.500.12573/418

Browse

Now showing 1 - 20 of 83

Systematic design optimization and implementation of line-start synchronous reluctance machine for down-hole submersible water pump applications
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2022) Tekgün, Didem; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
Considering the electric drive systems constitute roughly 40% of global energy production, improving electric machine efficiencies provides important nationwide and global scale advantages. Among the electric motors used in the industry, a major portion of them are pump motors used for pumping underground waters and petroleum products. Especially the motors for submersible pump applications run at very low-efficiency levels because of the motor design issues and wrong selection of motor-pump configurations. Due to the features like robustness, low cost, and line start capability, induction machines (IM) are generally the first choice for pump applications. However, IMs work with low efficiency, especially at low and medium power levels. Line start synchronous reluctance machines (LS-SynRM) come to the scene as a reasonable alternative by having the line start capability and not having rare earth permanent magnets as well. The working principle of these machines is a combination of a reluctance machine and an IM. In LS-SynRM, a rotor cage is inserted in the rotor for the machine to start with the line voltage, but the rotor copper losses become zero when the machine operates at synchronous speed. Moreover, SynRMs have higher power and torque density. In this thesis study, it is aimed to reduce the overall cost of the submersible water pump system by designing and optimizing a LS-SynRM as a submersible pump motor with higher efficiency compared to conventional IMs. Increasing the efficiency of the pump motor used in industry will improve the overall system performance. Accordingly, it lowers energy and maintenance costs, and easy process control will be achieved. This way, while reducing energy consumption nationwide significantly, not only the natural resources will be protected, but also huge amounts of money will be saved.
High-energy cosmic and gamma radiation measurement with remote-controlled secondary emission ionization calorimetry modules
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Paran, Nejdet; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
The demand for precise, robust, and reliable radiation-resistant particle detectors and ionization calorimeters intensifies, due to the escalating luminosity and unprecedented radiation conditions at particle colliders and accelerators. Secondary Emission (SE) Ionization Calorimetry is a novel technology designed to measure the energy of electromagnetic and hadronic particles, particularly in extreme radiation conditions. In this study, we have tested and investigated the development and radiation test of the novel SE modules. The modules were developed by modifying the conventional Hamamatsu single anode R7761 Photomultiplier Tubes. Three different voltage conditions for the same module were developed and the new modules were tested by using cosmic, gamma (Co-60) and neutron (AmBe) radiation sources. The results show that all three modes have good sensitivity to electromagnetic showers, and they are suitable for harsh radiation environments. This study also shows that SE module is a promising technology shedding light on future radiation-resistant nuclear and high-energy detectors. Here, we discuss the technical design, test characteristics and cosmic and particle interaction results of the newly developed SE modules. Since such detector systems are either in a high radiation area or in a closed room/box, remote mode changes allow us to continue the experimental process without interruption. By adding these signals to the interface where the modes are controlled, we can instantaneously observe the modes' effects.
Design and development of machine learning models for disease prediction and biomarkers detection
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Temiz, Mustafa; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
In medical science, the prediction of diseases and the identification of biomarkers play an important role in the diagnosis and treatment of various health conditions. The recent proliferation of data mining techniques has accelerated the development of disease prediction systems. In particular, machine learning methods are an effective way to analyze medical data and identify patterns to predict the likelihood of the disease development. Machine learning methods also help to identify biomarkers. Recently, the increasing incidence and mortality rates of inflammatory bowel disease, colorectal cancer and type 2 diabetes have drawn researchers' attention to these research areas. The aim of this thesis is to reduce the number of features and improve the prediction performance of machine learning based on complex biological datasets with a large number of disease-related features, as well as to identify potential biomarkers. In this thesis, three different studies are presented. The first study predicts eleven different cancer subgroups using miRNA data and biological domain knowledge and identifies potential biomarkers for these diseases. The second study predicts three different diseases using metagenomic data and biological domain knowledge and identifies potential biomarkers. The third study uses metagenomic data related to colorectal cancer to conduct global and population-based comprehensive experiments with traditional feature selection methods to identify potential biomarkers. This thesis presents a promising avenue for early disease detection, facilitating expedited treatment protocols, improving human survival rates, and potentially alleviating economic burdens within these critical research domains.
Enhancing breast cancer detection with a hybrid machine learning approach
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Etcil, Mustafa; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
According to the World Health Organization (WHO), breast cancer is one of the most prevalent illnesses, with 7.8 million instances recorded in the previous five years. As such, it poses a serious threat to world health. This alarming statistic underscores the urgent necessity for enhanced diagnostic methods. Against this backdrop, the current study proposes a novel diagnostic model, the CSA-PSO-LR classifier, which innovatively combines the clonal selection algorithm (CSA) with particle swarm optimization (PSO) to refine the logistic regression model training process for breast cancer detection. This research employs two extensively recognized datasets: the Wisconsin Diagnostic Breast Cancer (WDBC) and the Wisconsin Breast Cancer Database (WBCD), putting into practice a strict evaluation procedure that assesses performance using Bayesian hyperparameter optimization and 10-fold cross-validation. Furthermore, the study introduces CPU parallelization strategies to significantly curtail the model training time. Comparative analyses against machine learning algorithms, encompassing decision trees, extreme gradient boosting, k-nearest neighbors, logistic regression, random forests, and support vector machines, demonstrate the CSA-PSO-LR classifier's superior performance in detection accuracy and F1-measure. This investigation contributes a groundbreaking approach to the early detection of breast cancer, potentially facilitating more effective treatment plans and enhancing patient survival prospects.
Development of a machine learning-based system to identify disease biomarkers from human gut microbiota
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Koçak, Ayşegül; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
The human gut microbiota consists of a diverse ecosystem of organisms, encompasses billions of species. Recently developed next-generation sequencing methods have enabled researchers to examine the microbiota in greater detail, leading to new insights into its functions and dysfunctions. This study aims to identify metagenomic biomarkers (Microorganism-Enzyme Pairs) for colorectal cancer (CRC). The tool that we used allows for the analysis of microorganisms and enzymes within the gut microbiota. It achieves this by initially clustering enzymes based on their correlations with species and subsequently utilizing these clustering results to evaluate the ability of groups to differentiate between patient and healthy cohorts. By integrating species and enzymes, it is possible to identify pathogen microorganisms and enzyme clusters, that have the potential to distinguish cases (individuals with CRC) from controls (healthy individuals). The identified enzyme clusters and associated species could potentially act as biomarkers for colorectal cancer (CRC), enabling early diagnosis and more effective treatment. This approach holds promise for further exploration of the gut microbiota and its importance in human health and illness.
Tree-net: Bottleneck feature supervised network for biomedical image segmentation
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Demirci, Orhan; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
In this thesis, we introduce Tree-NET, a novel approach for medical image segmentation utilizing bottleneck feature supervision. This method enhances traditional segmentation algorithms by keeping supervision between bottleneck features of the network. The primary goal is to improve the model's ability to learn discriminative and robust features while simultaneously reducing computational costs. Bottleneck feature supervision involves compressing the input and label data using Autoencoders and then supervising the bottleneck features with a segmentation network named "Bridge-Net," which can be any segmentation model of choice. We applied Tree-NET to two critical medical image segmentation tasks: skin lesion segmentation and polyp segmentation. Our experiments demonstrate significant improvements in segmentation accuracy and efficiency. For instance, the U-NET backboned Tree-NET uses only 154.43 MB for executing and storing the model, which is almost 3.5 times smaller than the original U-Net while having a close number of trainable parameters. In skin lesion segmentation, Tree-NET achieved dice, Intersection-over-Union (IoU), and accuracy scores of 0.893, 0.751, and 0.977 respectively. For polyp segmentation, the scores were 0.856, 0.795, and 0.923 for dice, IoU, and accuracy respectively. Compared to traditional segmentation models, the empirical results show that Tree-NET achieves higher accuracy with reduced training time and computational cost, thus representing a significant advancement in medical image analysis by providing more reliable and efficient tools for clinical applications.
Machine learning based network anomaly detection
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Hacılar, Hilal; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
Intelligent technologies have led to a significant rise in internet users and applications. However, this rise in internet usage has also brought serious security challenges. Organizations rely on Network Intrusion Detection systems (NIDS) to protect sensitive data from unauthorized access and theft. To enhance the capabilities of IDS, Machine Learning (ML) and Deep Learning (DL) techniques have been increasingly integrated into these systems. In this context, anomaly-based network intrusion detection surpasses other detection mechanisms significantly in several instances. These systems analyze network traffic to detect suspicious activities, such as attempted breaches or cyberattacks. However, existing studies lack a thorough assessment of class imbalances, feature selection and extraction methods, hyperparameter optimization, and classification performance for different types of network intrusions: wired, wireless, and Software Defined Networking (SDN). Additionally, existing methods may achieve high accuracy; they may suffer from high training times, low detection rate (DR), and computational complexity. By combining metaheuristics and neural networks, it is possible to solve complex optimization problems that are challenging to solve using conventional methods. To address these challenges, this thesis study first evaluates different network intrusion datasets, such as wired, wireless, and SDN, together, considering class imbalance, feature selection, and hyperparameter optimization tasks. Secondly, it proposes a novel hybrid approach combining Deep Autoencoder (DAE) and Artificial Neural Network (ANN) models trained by a parallel Artificial Bee Colony (ABC) algorithm with Bayesian hyperparameter optimization.
Design and implementation of nanophotonic architectures using smart-self assembly of colloidal nanomaterials
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Şenel, Zeynep; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
DNA-driven self-assembly techniques offer precise control over the positioning of colloidal nanoparticles through specific Watson–Crick interactions, and its reversibility via controlling the temperature of medium. This thesis explores an alternative strategy to control DNA-functionalized nanoparticles' binding/unbinding process by leveraging laser radiation, inducing localized heating within the nanoparticles to facilitate disassociation. First, we demonstrate the active manipulation of the optical properties of DNA-assembled gold nanoparticle networks via external optical excitation. Specifically, irradiation with a green hand-held laser yields a substantial ∼30% increase in total transmittance, accompanied by a transition from opaque to transparent states observable in optical microscopy images. The reversibility of this process is demonstrated by the restoration of the nanoparticle network post-irradiation cessation, underscoring the efficacy of optical excitation in tailoring both the structure and optical characteristics of DNA-mediated nanoparticle assemblies. Second, we introduce a method to tailor DNA-driven self-assembly of semiconductor nanoparticles on glass by applying an external optical field. A green laser directs the assembly of DNA-functionalized red-emitting quantum dots (QDs) on DNA-functionalized glass, leaving uncoated spots owing to localized heating. This effect becomes prominent after three hours of radiation using a laser with an irradiance of 57.1 W/cm2. Experiments with different lasers and nanoparticle types confirm the role of laser-induced heating in preventing QD-glass bonding via DNA-DNA interaction. Secondary coating of previously uncoated spots with DNA-functionalized green-emitting QDs and dye-functionalized DNAs indicates a successful hierarchical self-assembly. Our findings highlight the potential of light-assisted DNA-driven self-assembly for diverse nanoparticle architectures, promising applications in optoelectronics and nanophotonics.
Design and performance enhancement of an ultra wideband vivaldi antenna
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Güzelkara, İzzet; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
Ultra-wideband technology has become a trending topic in the academic community since 2002 due to the release of the spectral mask by Federal Communications Commission, allowing the use of 3.6-10.1 GHz band for commercial and industrial applications. Being one of the fundamental components of ultra-wideband systems, ultra-wideband antennas are an important research area. In this research, Vivaldi antennas for ultra-wideband communications and several performance enhancement techniques for the antennas were studied. Antennas were designed and simulated using a commercially available three-dimensional electromagnetic simulation tool. First, a simple design of a Vivaldi antenna with a rectangular microstrip feed was obtained. The initial design has a -10 dB impedance bandwidth between 3.1 and 13.6 GHz and an average realized gain of 2.75 dBi. A method based on the alignment of the microstrip feed was described for adjusting the bandwidth of the initial design. Then, using several performance enhancement techniques such as implementation of corrugations and a parasitic patch, the antenna design was improved. Thanks to the applied methods, an antenna design with -10 dB impedance bandwidth extending from 1.33 to 10.1 GHz and an average realized gain of 6 dBi was achieved. Findings of this thesis study show that Vivaldi antennas having specific structures designed with the applied techniques are a promising solution for ultra-wideband communication systems, especially where antennas with directive radiation patterns are desired.
Decentralized electronic health record management system and disease prediction with machine learning methods
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Dedetürk, Beyhan Adanur; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
Electronic health records (EHRs) are vital to the advancement of healthcare and can help detect and prevent diseases early. However, EHR sharing faces challenges such as managing large data volumes, ensuring data privacy, security, and interoperability. This thesis aims to develop and analyze a blockchain-based EHR sharing system for disease prediction mechanism integration using SysML. The AguHyper platform, built by merging the InterPlanetary File System (IPFS) with Hyperledger Fabric, ensures the immutability of health records by storing hash values in the blockchain and encrypted records in IPFS. The system architecture and implementation configurations, including CouchDB and the Raft consensus mechanism, are thoroughly examined. The study also presents a novel hybrid approach called CSA-DE-LR, which integrates Differential Evolution (DE) and Clonal Selection Algorithm (CSA) with Logistic Regression (LR) to improve LR weights for precise categorization of cardiovascular diseases. The integration of the AguHyper with the CSA-DE-LR is explained in detail. At the end of our performance evaluations, we concluded that the AguHyper model has the potential to speed up the process of collecting and sharing data, and it offers an efficient platform for the participants.
Detection and classification of flaws from ultrasonic tomography images of composite materials based on deep learning
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Gülşen, Abdulkadir; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
This thesis introduces novel methodologies for enhancing defect classification and characterization in advanced composite materials by leveraging state-of-the-art machine learning (ML), deep learning (DL), and federated learning (FL) techniques within ultrasonic and acoustic emission (AE) inspection environments. First, a new ultrasonic dataset (UNDT), comprising 1,150 images from 60 distinct composite materials, is introduced. Applying transfer learning methods to both the UNDT and a publicly available dataset demonstrates the efficacy of advanced neural architectures—such as DenseNet121 and VGG19—achieving accuracy rates up to 98.8% and 98.6%, respectively. Next, the scope is extended to AE-based health monitoring by introducing an ensemble feature selection methodology to identify features strongly correlated with damage modes. By selecting amplitude and peak frequency for labeling and subsequently applying unsupervised clustering, the analysis confirms that both traditional AE features (e.g., counts and energy) and less commonly employed features (e.g., partial powers) correlate with distinct defect types. Finally, a novel FL framework is introduced to address the scarcity of publicly available, real-world ultrasonic datasets. This decentralized approach preserves data privacy while maintaining performance levels comparable to centralized methods, ensuring scalability and confidentiality in diverse data environments. Overall, these contributions significantly advance the field of NDT, offering robust defect classification and characterization. In doing so, the findings not only improve the accuracy and reliability of material integrity assessments but also lay a durable foundation for more secure, collaborative, and efficient NDT systems.
Time distributed classification of alzheimer's disease on MRI scans
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Dündar, Mehmet Sait; 0000-0002-0336-4825; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
This thesis presents a comprehensive framework for studying Alzheimer's Disease (AD) progression by focusing on the classification of AD, Mild Cognitive Impairment (MCI), and Cognitively Normal (CN) individuals using advanced machine learning models that analyze changes in brain volumetrics over time through MRI scans. In the first phase of the research, MR images from the Alzheimer's Disease Neuroimaging Initiative database were utilized, which included sequences of 3-4 scans taken annually from 22 CN, 18 AD, and 20 MCI subjects. Key volumetric parameters such as cortical thickness and intracranial volumes were extracted using the CAT12 toolbox in SPM software. A novel classification method based on the rate of volumetric changes over time was employed, effectively capturing the progressive nature of neurological changes. This approach achieved accuracies of 82.5% in distinguishing AD from CN, 71% in differentiating MCI from AD, and 69% in separating MCI from CN, alongside a 55% accuracy in a three-way classification using random forest and support vector machines. Building on these initial insights, the second phase of the study significantly advanced the methodology by integrating a pre-trained 3D ResNet 101 CNN algorithm for initial spatial categorization of MRI scans, followed by the use of Long Short-Term Memory (LSTM) networks. These LSTMs processed the same sequences of 3-4 annual scans for each patient, enhancing the model's ability to analyze and interpret the temporal progression of volumetric changes. This sophisticated approach led to marked improvements in classification accuracy: 96.7% in differentiating AD from CN, 87.5% in distinguishing AD from MCI, and 86.4% in separating MCI from CN. The study effectively demonstrates a significant enhancement in capturing the temporal dynamics of AD progression.
Tumor detection in breast cancer histopathological images using convolutional neural networks
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Şahbaz, Zeki; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
Breast cancer is one of the most common cancer types among women worldwide. Early detection significantly increases the chances of survival and effective treatment, making advancements in diagnostic methodologies crucial. This study aims to improve the detection of tumor cells in breast cancer histopathology images using deep learning and image processing techniques. Significant modifications have been made to the hyperparameters, including the tumor bounding box size, batch size, optimization algorithms, learning rate, and weight decay. These changes focus on determining the best parameters of the Faster R-CNN model. A comprehensive analysis of different parameters was conducted using the Breast Cancer Histopathological Annotation and Diagnosis (BreCaHAD) dataset. The analysis identified the best settings for model performance, shows by improvements in precision, recall, and F-score. Our research contributes to the field of medical image analysis by identifying critical factors that affect the accuracy of tumor detection, contributing to the development of more accurate diagnostic tools.
Perception estimation and torque control for hand prostheses using EEG and EMG signals
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Karakullukcu, Nedime; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
Upper extremity prostheses vary based on patients' articulation levels and movement methods. They can be cosmetic, operate mechanically with shoulder movement, or be controlled by myoelectronic and electroencephalography (EEG) signals. However, unnatural prosthesis control burdens users mentally. This thesis seeks to enhance bionic hand prosthesis control using EEG and electromyography (EMG) signals, coupled with users' visual weight perception, aiming to reduce physical and mental discomfort associated with mechanical prostheses. The prototype hand's preconditioning evaluates objects' weight visually, aiming to reduce shoulder force and mental load while holding an object. EEG and EMG signals from subjects were processed for real-time implementation. In the first stage, a study focused on operating the prosthesis using the motor intention waves of prosthesis users, and the machine learning approaches' classification success (detection of the intention to activate the prosthesis) was examined using EEG data from 30 healthy participants. The second stage recorded EEG and EMG signals from 31 participants during reaching, lifting, and placing an object, employing various classifications for object weight. In the real-time classification of multi-channel EEG signals from 20 healthy individuals using Fourier-based synchrosequeezing transform (FSST) and singular value decomposition (SVD) approaches, the system aimed to control the stiffness of the wrist part of the prosthesis. Consequently, the system could detect the weight of the object perceived by the user while using the prosthesis, allowing for the preconditioning of the prosthesis based on this weight when the user wants to hold and move the object.
Machine learning approaches for internet of things based vehicle type classification and network anomaly detection
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Kolukısa, Burak; 0000-0003-0423-4595; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
This thesis presents innovative approaches in the realms of Intelligent Transportation Systems (ITS) and Network Intrusion Detection Systems (NIDS) within the Internet of Things (IoT). Leveraging IoT technologies, a low-cost, battery-operated 3-D magnetic sensor has been developed for ITS to enable the classification of vehicle categories. The research presents machine learning and deep learning models that are improved by using oversampling, feature selection and extraction methods, hyperparameter optimization, and converting signals into 2-D images. New methods have been proposed for vehicle type classification to boost classification performance and achieve an accuracy of up to 92.92%. Additionally, the increasing reliance on IoT devices for such applications introduces significant cybersecurity risks. To mitigate these vulnerabilities, a novel logistic regression model trained with a parallel artificial bee colony (LR-ABC) algorithm has been proposed for network anomaly detection. This model incorporates hyperparameter optimization to enhance detection capabilities, showcasing superior performance on popular benchmark NIDS datasets with accuracies of 88.25% and 90.11%. Overall, this research contributes to the advancement of IoT and IoT cybersecurity by offering robust, scalable, and efficient solutions. These innovations not only enhance vehicle type classification and network security in the IoT era but also pave the way for future IoT infrastructure development in an increasingly connected digital landscape.
Analysing network traffic and detecting network threats by using the algorithms of machine learning
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Küçükkoç, Abdurrahman; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
As information technologies progress, the possibilities of access to information increase and therefore it becomes difficult to ensure the security of information. Today, with the use of information systems in all areas of life, network threats have also increased. The increase in individual access to and use of the internet has also brought network threats. In addition, the latest developments in information technologies, developing global communication networks, the internet of things aiming to connect all objects with networks, cloud technologies, the spread of mobile internet and the renewal of devices have brought network threats and uncertainties. Network threats increase the security vulnerabilities in the information and communication systems of individuals and organisations day by day. This situation causes systems to malfunction, economic damage and cyber security to be jeopardised. In order to contribute to individuals, institutions and organisations, our thesis aims to protect information systems against network threats, to ensure data confidentiality, integrity and accessibility, to detect network threats in advance and to take measures against these threats. We believe that by analysing heterogeneous network traffic, which includes most network attacks on the Internet, and using machine learning algorithms, we will reach a result close to reality in the detection of network threats. In line with this result, we will be able to take precautions against network threats in information systems and structures
Machine learning methods for detecting genetic and infectious diseases
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Işık, Yunus Emre; 0000-0001-6176-7545; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
Completion of the whole human genome in the 2003 has led to various advances in many fields, particularly in biology, genetics, health sciences, treatment, and pharmacology. In the following years, spread of faster and cheaper sequencing technologies has enabled us to extract and analyze genetic profiles of individuals digitally. Consequently, individual-specific forecasting and personalized treatment and precision medicine-, what once seemed like science fiction, have become more and more real. In both approaches, one of the crucial steps is identifying the presence of diseases using individual-specific genetic data. This thesis aims to comprehensively and comparatively evaluate the predictive performance of machine learning methods for Behçet's disease and respiratory infections. Additionally, feature selection methods were employed to identify the genetic factors (such as SNPs and genes) associated with disease presence for both diseases. Furthermore, the usability of selected features depending on biological pathway-driven active subnetworks listed in the literature was analyzed for the prediction of Behçet's disease. For the respiratory infection prediction problem, on the other hand, the prediction performance of features calculated by single-sample gene set enrichment analysis (ssGSEA) was evaluated using different machine learning methods. As the data types used in both experiments were different (genome-wide association studies data, gene expression profiles), the performance of machine learning approaches on different data types was also observed. It is hoped that the findings of both experiments will contribute to future machine learning based disease prediction studies.
Deep learning models for traffic volume prediction
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Çini, Nevin; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
In the last 50 years, with the growth of cities and increase in the number of vehicles and mobility, traffic has become troublesome. As a result, traffic flow prediction started to attract attention as an important research area. However, despite the extensive literature, traffic flow prediction still remains as an open research problem, specifically for long- term traffic flow prediction. Compared to the models developed for short-term traffic flow prediction, the number of models developed for long-term traffic flow prediction is very few. Based on this shortcoming, in this study, we focus on long-term traffic flow prediction and propose a novel deep ensemble model (DEM). In order to build this ensemble model, first, we developed a convolutional neural network (CNN), a long short term memory (LSTM) network, and a gated recurrent unit (GRU) network as deep learning models, which formed the base learners. In the next step, we combine the output of these models according to their individual forecasting success. We use another deep learning model to determine the success of the individual models. Our proposed model is a flexible ensemble prediction model that can be updated based on traffic data. To evaluate the performance of the proposed model, we use a publicly available dataset. Numerical results show that our proposed model performs better than individual deep learning models (i.e., LSTM, CNN, GRU), selected traditional machine learning models (i.e., linear regression (LR), decision tree regression (DTR), k-nearest-neighbors regression (KNNR) and other ensemble models such as random-forest-regression(RFR).
Diagnosis of coronary artery disease with machine learning approaches
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Halıcı, İkram; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
The World Health Organization states that Coronary Artery Disease (CAD) ranks as a primary cause of recorded fatalities. CAD occurs as a result of the blockage of coronary artery vessels, which are located on the surface of the heart and supply the blood that the heart needs. Diagnosing the disease using traditional methods is challenging and requires costly tests. In recent years, the use of machine learning-based methods has increased as an alternative diagnostic approach. However, existing studies in the literature suffer from low detection rates and long training times. Therefore, there is still a need for reliable and low-cost diagnostic methods. In this thesis, a new model, CSA-PSO-ANN, is proposed for the diagnosis of coronary artery disease. The aim is to reduce the training time of the machine learning model and achieve a higher accuracy in diagnosing the disease. Experiments have been conducted on two publicly available datasets. Parallelization, feature selection, and hyperparameter optimization have been performed to shorten the model's training time. The performance of the model has been compared with well-known machine-learning algorithms and previous studies. The experiments showed that the proposed model effectively diagnoses the disease and outperforms other methods in terms of accuracy and F1 score performance metrics.
Investigation and improvement of the smooth mode transition technique for quasi-single-stage four-switch buck-boost inverter
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2023) Keskinkılıç, Ebubekir; 0000-0002-4913-6684; AGÜ, Fen Bilimleri Enstitüsü, Elektrik ve Bilgisayar Mühendisliği Ana Bilim Dalı
In recent decades, given the world's inevitable energy scarcity, increasing energy demand and green energy concerns, high efficiency energy conversion has become more important and attractive than ever, and researchers have directed their interest to energyefficient converters. Inverters are a commonly utilized type of converter, which can be classified into two categories: single and two-stage inverters. Considering the inherent drawbacks of traditional inverters, a quasi-single-stage inverter (QSSI) has emerged. The QSSI uses a DC-DC converter to shape the rectified version of the desired AC waveform in the first stage and, in the second stage, it switches only once to alternate the polarity. It stands forward in terms of efficiency, control simplicity, and system stability. Among QSSI, a non-inverting buck-boost converter has drawn attention due to its capability to perform both step-up and down modes and its bidirectional power transfer feature. In the first stage of the QSS non-inverting buck-boost converter; smooth transitions between the buck and boost modes and efficient conversion cannot be achieved by the traditional two-mode control method when the output voltage level is close to the input voltage level due to various limitations, non-idealities, and disturbances. Many methods have been applied and studied in the literature to minimize or eliminate the effects of the region which is called the “dead zone”. In this thesis study, further efficiency and THD improvement for the QSSI is targeted by employing a four-mode control method. The study incorporates a comparative study of the dead zone effects on inverter systems, which have not been previously documented in the literature. Moreover, it places a priority on optimizing efficiency and minimizing distortion in various applications— ranging from motor control and solar energy systems to grid-tied wind turbines and switched-mode power supplies—by comparing existing methods with open-loop voltage control. In conclusion, the theoretical results are verified with experimental studies.

Browse

Recent Submissions