Browsing by Author "Özçevik Y."
Now showing 1 - 11 of 11
Results Per Page
Sort Options
Item Energy-aware mobility for aerial networks: A reinforcement learning approach(John Wiley and Sons Ltd, 2022) Bozkaya E.; Özçevik Y.; Akkoç M.; Erol M.R.; Canberk B.With recent advancements in aerial networks, aerial base stations (ABSs) have become a promising mobile network technology to enhance the coverage and capacity of the cellular networks. ABS deployment can assist cellular networks to support network infrastructure or minimize the disruptions caused by unexpected and temporary situations. However, with 3D ABS placement, the continuity of the service has increased the challenge of providing satisfactory Quality of Service (QoS). The limited battery capacity of ABSs and continuous movement of users result in frequent interruptions. Although aerial networks provide quick and effective coverage, ABS deployment is challenging due to the user mobility, increased interference, handover delay, and handover failure. In addition, once an ABS is deployed, an intelligent management must be applied. In this paper, we model user mobility pattern and formulate energy-aware ABS deployment problem with a goal of minimizing energy consumption and handover delay. To this end, the contributions of this paper are threefold: (i) analysis of reinforcement learning (RL)-based state action reward state action (SARSA) algorithm to deploy ABSs with an energy consumption model, (ii) predicting the user next-place with a hidden Markov model (HMM), and (iii) managing the dynamic movement of ABSs with a handover procedure. Our model is validated by comprehensive simulation, and the results indicate superiority of the proposed model on deploying multiple ABSs to provide the communication coverage. © 2021 John Wiley & Sons, Ltd.Item MetricHunter: A software metric dataset generator utilizing SourceMonitor upon public GitHub repositories(Elsevier B.V., 2023) Özçevik Y.; Altay O.Version control systems are pervasively consulted nowadays to obtain software metric datasets. Accordingly, machine learning is applied to predict different aspects of a software including quality monitoring, influence analysis, etc. However, construction of a metric dataset is challenging and the dataset content may affect the success of the learning-based models. In this study, we propose a dataset construction tool, MetricHunter, which is able to produce platform/language specific datasets that can be used for predicting the features of newly created software. The proposed tool is developed by C# programming language utilizing a known metric gathering tool, i.e. SourceMonitor, and the GitHub REST API for public repositories. Thus, one can construct a proper dataset from a graphical user interface by simply specifying the programming language or target platform. The outputs of the tool on a set of repositories are validated by investigating automatically generated attribute values and comparing them with the measurements of metric gathering tools as well as the GitHub metric values. © 2023 The Author(s)Item Human robot interaction as a service for combatting COVID-19: an experimental case study(Springer Science and Business Media Deutschland GmbH, 2023) Özçevik Y.COVID-19 pandemic has changed today’s routines in a variety of fields such as society, economics, health, etc. It is surely known that the most powerful weapon to fight against the disease is the social distancing. Hence, it is strongly recommended by the authorities to decrease human to human interaction (HHI) in order to stop the spread. However, daily routine of people must continue somehow, because of the fact that it is not known when the pandemic will end permanently. Thus, new approaches should be adapted in social environments for COVID-19 prevention. Human robot interaction (HRI) can be seen as a vital mechanism to provide risk free routines in the society. For this purpose, we offer a human robot interaction as a service (HRIaaS) for eatery locations such as restaurants, cafes, etc. where customers should interact with the staff. The proposed service aims to utilize personal smartphones and decrease the number of HHIs for such environments in which strange people involved. Moreover, an experimental case study is conducted to obtain an evaluation with a real world scenario when the proposed service is used versus a contemporary routine with HHIs. The evaluation results show that an average reduction of 41% is achieved per customer in the number of HHIs between customers and serving staff. © 2022, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.Item A Real-Time Nut-Type Classifier Application Using Transfer Learning(Multidisciplinary Digital Publishing Institute (MDPI), 2023) Özçevik Y.Smart environments need artificial intelligence (AI) at the moment and will likely utilize AI in the foreseeable future. Shopping has recently been seen as an environment needing to be digitized, especially for payment processes of both packaged and unpackaged products. In particular, for unpackaged nuts, machine learning models are applied to newly collected dataset to identify the type. Furthermore, transfer learning (TL) has been identified as a promising method to diminish the time and effort for obtaining learning models for different classification problems. There are common TL architectures that can be used to transfer learned knowledge between different problem domains. In this study, TL architectures including ResNet, EfficientNet, Inception, and MobileNet were used to obtain a practical nut-type identifier application to satisfy the challenges of implementing a classifier for unpackaged products. In addition to the TL models, we trained a convolutional neural network (CNN) model on a dataset including 1250 images of 5 different nut types prepared from online-available and manually captured images. The models are evaluated according to a set of parameters including validation loss, validation accuracy, and F1-score. According to the evaluation results, TL models show a promising performance with 96% validation accuracy. © 2023 by the author.Item Proof of Evaluation-based energy and delay aware computation offloading for Digital Twin Edge Network(Elsevier B.V., 2023) Bozkaya E.; Erel-Özçevik M.; Bilen T.; Özçevik Y.The increasing availability of Internet of Things (IoT) applications has led to the development of new technologies. Specifically, the deployment of edge servers close to IoT devices has strengthened the edge computing paradigm. With the collaboration of Mobile Edge Computing (MEC) and cloud computing, delay-sensitive and computation-intensive tasks can be offloaded to the edge/cloud servers to improve system performance in terms of the delay and energy consumption of IoT devices. However, there is a need to schedule the computation tasks for an efficient management. More importantly, the task scheduling strategy can face data tampering attacks to deliberately modify, destroy or manipulate the decisions. To solve the above problems, in this paper, we newly propose to integrate digital twin and blockchain into the edge networks. However, it is unclear (i) how energy and delay-aware computation should be combined, and (ii) which mining computations should be executed for a secure task scheduling. The state-of-the-art focuses on task scheduling and blockchain mining, separately. Therefore, we propose a novel blockchain-based digital twin-edge network architecture where our proposed algorithm solves these two challenges at the same execution. We design a three-layer system architecture, composed of physical entity layer, digital twin edge layer and blockchain layer. In the physical entity layer, we formulate an energy and delay-aware task scheduling problem. In the digital twin edge layer, we propose a novel Proof of Evaluation (PoE)-based secure energy and delay-aware task scheduling algorithm where optimization is executed by the genetic algorithm implementation of Warehouse Location Problem (WLP). In the blockchain layer, the best-found solutions are shared with the topology in a blockchain. Here, each block includes the hash of the previous block, a genetic algorithm-based solution, nonce value, and a hash of whole blocks in the blockchain. Thus, we aim to execute the computation tasks with an acceptable delay in an energy-efficient manner and prevent data tampering attacks against the optimal computation decisions. We validate the outcomes of our PoE-based secure digital twin-edge network model with extensive evaluations. Since the proposed model distributes the task not only to the local device but also to the MEC and cloud server for delay awareness, it reduces the delay but consumes more energy. Nevertheless, the additional energy consumption can be neglected against the delay reduction. The proposed scheme is also more scalable to compare with the conventional solution. The numerical results clearly show that the proposed model provides energy and delay awareness, maintaining both data integrity and trustworthiness at the same execution of algorithm. © 2023 Elsevier B.V.Item A Comparative Study of Metaheuristic Optimization Algorithms for Solving Real-World Engineering Design Problems(Tech Science Press, 2023) Altay E.V.; Altay O.; Özçevik Y.Real-world engineering design problems with complex objective functions under some constraints are relatively difficult problems to solve. Such design problems are widely experienced in many engineering fields, such as industry, automotive, construction, machinery, and interdisciplinary research. However, there are established optimization techniques that have shown effectiveness in addressing these types of issues. This research paper gives a comparative study of the implementation of seventeen new metaheuristic methods in order to optimize twelve distinct engineering design issues. The algorithms used in the study are listed as: transient search optimization (TSO), equilibrium optimizer (EO), grey wolf optimizer (GWO), moth-flame optimization (MFO), whale optimization algorithm (WOA), slime mould algorithm (SMA), harris hawks optimization (HHO), chimp optimization algorithm (COA), coot optimization algorithm (COOT), multi-verse optimization (MVO), arithmetic optimization algorithm (AOA), aquila optimizer (AO), sine cosine algorithm (SCA), smell agent optimization (SAO), and seagull optimization algorithm (SOA), pelican optimization algorithm (POA), and coati optimization algorithm (CA). As far as we know, there is no comparative analysis of recent and popular methods against the concrete conditions of real-world engineering problems. Hence, a remarkable research guideline is presented in the study for researchers working in the fields of engineering and artificial intelligence, especially when applying the optimization methods that have emerged recently. Future research can rely on this work for a literature search on comparisons of metaheuristic optimization methods in real-world problems under similar conditions. © 2023 Tech Science Press. All rights reserved.Item A real-time simulation environment architecture for autonomous vehicle design; [Otonom araç tasarimi için gerçek zamanli benzetim ortami mimarisi](Gazi Universitesi, 2023) Özçevik Y.; Solmaz Ö.; Baysal E.; Ökten M.Various proposed approaches for autonomous driving basically involve an image processing and a machine learning process. It is extremely important to use appropriate image processing techniques and a comprehensive data set in these approaches. Moreover, the proposed model must work in real-time. On the other hand, designing and manufacturing an autonomous vehicle model results in serious hardware costs. In addition, the design and manufacturing processes need to be repeated to develop new approaches. In this context, utilizing a real-time simulation environment can be seen as a suitable approach for a less costly prevalidation of such models. In this study, a real-time simulation architecture is developed with Unity framework to test an autonomous driving model. In addition, an autonomous driving model that includes lane tracking and object recognition approaches is proposed, and an autonomous vehicle simulation is created. Finally, the feasibility of the proposed simulation architecture is tested with the convolutional neural networks-based YOLO algorithm and R-CNN algorithm versions. According to the findings, it is observed that Faster R-CNN, Mask R-CNN and YOLO-v4 algorithms produce results with 91%, 93% and 95% accuracy, respectively. It has been determined that these results are close to the accuracy rates obtained on different traffic sign data sets in the literature. Considering the outcomes, it is argued that a vehicle simulation with an autonomous driving model has been successfully tested in the proposed system architecture. © 2023 Gazi Universitesi Muhendislik-Mimarlik. All rights reserved.Item A Genetic Optimized Federated Learning Approach for Joint Consideration of End-to-End Delay and Data Privacy in Vehicular Networks(Multidisciplinary Digital Publishing Institute (MDPI), 2024) Erel-Özçevik M.; Özçift A.; Özçevik Y.; Yücalar F.In 5G vehicular networks, two key challenges have become apparent, including end-to-end delay minimization and data privacy. Learning-based approaches have been used to alleviate these, either by predicting delay or protecting privacy. Traditional approaches train machine learning models on local devices or cloud servers, each with their own trade-offs. While pure-federated learning protects privacy, it sacrifices delay prediction performance. In contrast, centralized training improves delay prediction but violates privacy. Existing studies in the literature overlook the effect of training location on delay prediction and data privacy. To address both issues, we propose a novel genetic algorithm optimized federated learning (GAoFL) approach in which end-to-end delay prediction and data privacy are jointly considered to obtain an optimal solution. For this purpose, we analytically define a novel end-to-end delay formula and data privacy metrics. Accordingly, a novel fitness function is formulated to optimize both the location of training model and data privacy. In conclusion, according to the evaluation results, it can be advocated that the outcomes of the study highlight that training location significantly affects privacy and performance. Moreover, it can be claimed that the proposed GAoFL improves data privacy compared to centralized learning while achieving better delay prediction than other federated methods, offering a valuable solution for 5G vehicular computing. © 2024 by the authors.Item Data-oriented QMOOD model for quality assessment of multi-client software applications(Elsevier B.V., 2024) Özçevik Y.There has been a great effort to evaluate software quality using proper tools and methods against different development environments changing over time. Quality Model for Object Oriented Design (QMOOD) is a verified model used for quality assessment of object-oriented software. The model associates quality metrics gathered from the source code and quality attributes in use to present a quality measurement. However, the model should be revised for recent multi-client software including native client applications, because there is a deficiency of metric gathering tools in such environments. More specifically, it is sometimes not possible to gather all quality properties required by QMOOD in all native development platforms of client applications. Hence, even though different client applications have the same design, the implementation quality cannot be monitored for the quality assurance. Analyzing and simplifying the metric set may alleviate this challenge, and a convenient quality assessment might be achieved. Thus, we propose to change the operational aspect of QMOOD by inserting an additional layer, Data Analytic, to the hierarchical structure of the conventional model. Accordingly, we provide a discussion on a case study including five native client applications. For this purpose, a design quality of one of the client applications is achieved to validate the appropriateness of the design, the data analytic on the metric set are implemented and the proposed data-oriented simplified QMOOD is applied to the other client applications. Finally, it is stated that the proposed approach successfully alleviated the problems in metric gathering for multi-client applications while applying QMOOD. © 2024Item Average Localization Error Prediction for 5G Networks: An Investigation of Different Machine Learning Algorithms(Springer, 2024) Altay O.; Erel-Özçevik M.; Varol Altay E.; Özçevik Y.In the realm of today’s networking technologies, user localization has been a formidable challenge for recent applications. There are different approaches in pursuit of heightened position detection of an end-user with the help of GPS, Wi-Fi fingerprint and 5G equipment. However, these approaches require both deployment and maintenance costs because of equipment establishment for position tracking. Moreover, they are not capable of minimizing the localization error, especially for indoor scenarios to track the indoor position of an end-user. Hence, there is an urgent need to delve deeper into innovative approaches to drive further advancements in user localization. In response, Machine Learning (ML) approaches have recently been widely adapted to predict the localization of end-users with minimum error. More specifically, average localization error (ALE) of an end-user can be predicted in a cost-effective way by using proper data and ML methods. For this purpose, we have investigated different ML approaches to get an accurate ALE prediction scheme for 5G networks with mobile end-users. Accordingly, an existing dataset is utilized to generate localization data of end-users in which the ALE is directly calculated by Received Signal Strength Indicator. Moreover, three different normalization approaches are applied for the overarching goal of increased data quality. Consequently, six different ML algorithms, including Linear regression, support vector machine with three different kernels, Gaussian process, and ensemble least-squares boosting (LSBoost) are evaluated with respect to a set of evaluation criteria including R, R2, RMSE, and MAE. The evaluation outcomes emphasize that ensemble LSBoost method, in the context of localization prediction, outperforms the other approaches and is sufficient to yield a viable learning strategy for ALE prediction. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.Item An embedded TensorFlow lite model for classification of chip images with respect to chip morphology depending on varying feed(Springer, 2025) Özçevik Y.; Sönmez F.Turning is one of the fundamental machining processes used to produce superior machine parts. It is critical to manage the machining conditions to maintain the desired properties of the final product. Chip morphology and chip control are crucial factors to be monitored. In particular, the selection of an appropriate feed has one of the most significant effects. On the other hand, machine learning is an advanced approach that is continuously evolving and helping many industries. Moreover, mobile applications with learning models have been deployed in the field, recently. Taking these motivations into account, in this study, we propose a practical mobile application that includes an embedded learning model to provide chip classification based on chip morphology. For this purpose, a dataset of chips with different morphological properties is obtained and manually labeled according to ISO 3685 standards by using 20 different feeds on AISI 4140 material. Accordingly, TensorFlow Lite is used to train a learning model, and the model is embedded into a real-time Android mobile application. Eventually, the final software is evaluated through experiments conducted on the dataset and in the field, respectively. According to the evaluation results, it can be stated that the learning model is able to predict chip morphology with a test accuracy of 85.4%. Moreover, the findings obtained from the real-time mobile application satisfy the success rate by practical usage. As a result, it can be concluded that such attempts can be utilized in the turning process to adjust the relevant feed conditions. © The Author(s) 2024.