Scopus İndeksli Yayınlar Koleksiyonu
Permanent URI for this collectionhttps://hdl.handle.net/20.500.12573/395
Browse
9 results
Search Results
Conference Object Citation - Scopus: 1Traffic Light Management Systems Using Reinforcement Learning(Institute of Electrical and Electronics Engineers Inc., 2022-09-07) Can, Sultan Kubra; Thahir, Adam Rizvi; Cos¸kun, Mustafa; Güngör, Vehbi Çağrı; Coskun, MustafaWhile reducing traffic congestion and decrease the number of traffic accidents in the intersections, most of the traffic light management approaches cannot adapt well to fast changing traffic dynamics and growing demands of the intersections with modern world developments. To overcome this problem, adaptive traffic controllers are developed, and detectors and sensors are added to systems to enable adoption and dynamism. Recently, reinforcement learning has shown its capability to learn the dynamics of complex environments, such as urban traffic. Although it was studied in single junction systems, one of the problems was the lack of consistency with how the real world system works. Most of the systems assume that the environment is fully observable or actions would be freely executed using simulators. This study aims to merge usefulness of reinforcement learning methods with real-world traffic constraints. Comparative performance evaluations show that the reinforcement learning algorithm (Advantage Actor-Critic (A2C)) converges well while staying stable under changing traffic dynamics. © 2022 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 4Re-Exploring the Kayseri Culture Route by Using Deep Learning for Cultural Heritage Image Classification Cultural Heritage Image Classification by Using Deep Learning: Kayseri Culture Route(Association for Computing Machinery, 2024-05-25) Kevseroğlu, Ozlem; Kurban, RifatThe categorization of images captured during the documentation of architectural structures is a crucial aspect of preserving cultural heritage in digital form. Dealing with a large volume of images makes this categorization process laborious and time-consuming, often leading to errors. Introducing automatic techniques to aid in sorting would streamline this process, enhancing the efficiency of digital documentation. Proper classification of these images facilitates improved organization and more effective searches using specific terms, thereby aiding in the analysis and interpretation of the heritage asset. This study primarily focuses on applying deep learning techniques, specifically SqueezeNet convolutional neural networks (CNNs), for classifying images of architectural heritage. The effectiveness of training these networks from scratch versus fine-tuning pre-existing models is examined. In this study, we concentrate on identifying significant elements within images of buildings with architectural heritage significance of Kayseri Culture Route. Since no suitable datasets for network training were found, a new dataset was created. Transfer learning enables the use of pre-trained convolutional neural networks to specific image classification tasks. In the experiments, 99.8% of classification accuracy have been achieved by using SqueezeNet, suggesting that the implementation of the technique can substantially enhance the digital documentation of architectural heritage. © 2024 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 8Generating Emergency Evacuation Route Directions Based on Crowd Simulations With Reinforcement Learning(Institute of Electrical and Electronics Engineers Inc., 2022-09-07) Unal, Ahmet Emin; Gezer, Cengiz; Kuleli Pak, Burcu Kuleli; Güngör, Vehbi Çağrı; Pak, Burcu KuleliIn an emergency, it is vital to evacuate individuals from the dangerous environments. Emergency evacuation plan-ning ensures that the evacuation is safe and optimal in terms of evacuation time for all of the people in evacuation. To this end, the computer-enabled evacuation simulation systems are used to generate optimal routes for the evacuees. In this paper, a dynamic emergency evacuation route generator has been proposed based on indoor plans of the building and the locations of the evacuees. To generate the optimal routes in real-time, a reinforcement learning algorithm (proximal policy optimization) is presented. Comparative performance results show that the proposed model is successful for evacuating the individuals from the building in different scenarios. © 2022 Elsevier B.V., All rights reserved.Article Citation - Scopus: 14CoviDetector: A Transfer Learning-Based Semi Supervised Approach to Detect COVID-19 Using CXR Images(Elsevier B.V., 2023-06) Chowdhury, Deepraj; Das, Anik; Dey, Ajoy; Banerjee, Soham; Golec, Muhammed; Kollias, Dimitrios; Arya, Rajesh Chand; Uhlig, SteveCOVID-19 was one of the deadliest and most infectious illnesses of this century. Research has been done to decrease pandemic deaths and slow down its spread. COVID-19 detection investigations have utilised Chest X-ray (CXR) images with deep learning techniques with its sensitivity in identifying pneumonic alterations. However, CXR images are not publicly available due to users’ privacy concerns, resulting in a challenge to train a highly accurate deep learning model from scratch. Therefore, we proposed CoviDetector, a new semi-supervised approach based on transfer learning and clustering, which displays improved performance and requires less training data. CXR images are given as input to this model, and individuals are categorised into three classes: (1) COVID-19 positive; (2) Viral pneumonia; and (3) Normal. The performance of CoviDetector has been evaluated on four different datasets, achieving over 99% accuracy on them. Additionally, we generate heatmaps utilising Grad-CAM and overlay them on the CXR images to present the highlighted areas that were deciding factors in detecting COVID-19. Finally, we developed an Android app to offer a user-friendly interface. We release the code, datasets and results’ scripts of CoviDetector for reproducibility purposes; they are available at: https://github.com/dasanik2001/CoviDetector © 2024 Elsevier B.V., All rights reserved.Article Citation - Scopus: 10Building a Challenging Medical Dataset for Comparative Evaluation of Classifier Capabilities(Elsevier Ltd, 2024-08) Bozkurt, Berat; Coskun, Kerem; Bakal, GokhanSince the 2000s, digitalization has been a crucial transformation in our lives. Nevertheless, digitalization brings a bulk of unstructured textual data to be processed, including articles, clinical records, web pages, and shared social media posts. As a critical analysis, the classification task classifies the given textual entities into correct categories. Categorizing documents from different domains is straightforward since the instances are unlikely to contain similar contexts. However, document classification in a single domain is more complicated due to sharing the same context. Thus, we aim to classify medical articles about four common cancer types (Leukemia, Non-Hodgkin Lymphoma, Bladder Cancer, and Thyroid Cancer) by constructing machine learning and deep learning models. We used 383,914 medical articles about four common cancer types collected by the PubMed API. To build classification models, we split the dataset into 70% as training, 20% as testing, and 10% as validation. We built widely used machine-learning (Logistic Regression, XGBoost, CatBoost, and Random Forest Classifiers) and modern deep-learning (convolutional neural networks - CNN, long short-term memory - LSTM, and gated recurrent unit - GRU) models. We computed the average classification performances (precision, recall, F-score) to evaluate the models over ten distinct dataset splits. The best-performing deep learning model(s) yielded a superior F1 score of 98%. However, traditional machine learning models also achieved reasonably high F1 scores, 95% for the worst-performing case. Ultimately, we constructed multiple models to classify articles, which compose a hard-to-classify dataset in the medical domain. © 2024 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 4A Transfer Learning Application on the Reliability of Psychological Drugs' Comments(Institute of Electrical and Electronics Engineers Inc., 2023-07-25) Sen, Tarik Uveys; Bakal, GokhanAs digitalization and the Internet stay emerging concepts by gaining popularity, the accuracy of personal reviews/opinions will be a critical issue. This circumstance also particularly applies to patients taking psychological drugs, where accurate information is crucial for other patients and medical professionals. In this study, we analyze drug reviews from drugs.com to determine the effectiveness of reviews for psychological drugs. Our dataset includes over 200,000 drug reviews, which we labeled as positive, negative, or neutral according to their rating scores. We apply machine learning (ML) models, including Logistic Regression, Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) algorithms, to predict the sentiment class of each review. Our results demonstrate an F1-Weighted score of 85.3% for the LSTM model. However, by applying the transfer learning technique, we further improved the F1 score (nearly 3% increase) obtained by the LSTM model. Our findings proved that there is no contextual difference between the comments made by the patients suffering from psychological or other diseases. © 2023 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 19A Novel Feature Design and Stacking Approach for Non-Technical Electricity Loss Detection(Institute of Electrical and Electronics Engineers Inc., 2018-05) Aydin, Zafer; Güngör, Vehbi ÇağrıNon-technical electricity losses continue to jeopardize economic and social well-being of many countries. In this work, we develop machine learning classifiers that can identify anomalous electricity consumption in Turkey. Starting from weekly electricity usage data, we develop new features that capture statistical and frequency domain characteristics of the customers and their consumption patterns. We analyze the effect of reducing number of feature descriptors through dimensionality reduction and feature selection techniques. To overcome the class imbalance problem, we implement several ensemble methods and compare their prediction accuracy to those of the standard classifiers. The proposed features and combining strengths of different classifiers bring significant improvements on performance metrics, which is demonstrated through detailed simulations on shopping mall sector. We anticipate that advances in this field will contribute to the economies considerably. © 2018 Elsevier B.V., All rights reserved.Conference Object Population Specific Classification of Colorectal Cancer With Meta-Analysis of Metagenomic Data(Institute of Electrical and Electronics Engineers Inc., 2023-10-11) Temiz, Mustafa; Yousef, Malik; Bakir-Güngör, BurcuAdvances in next-generation sequencing and '-omics' technologies makes it possible to characterize the human gut microbiome. While some of these microorganisms are important regulators of our immune system, modulation of the microbiota leads to a variety of diseases. Colorectal cancer (CRC), the third most common cancer worldwide, is caused by genetic mutations, environmental conditions, and abnormalities in the gut microbiota. Using various machine learning methods and meta-analysis techniques, this study aims to build a classification model that can help in CRC diagnosis by analyzing metagenomic datasets of different populations obtained at the species level. Using 8 different countries and 9 different metagenomic datasets, 3 different meta-analyzes are performed: within-population, cross-population, and one population is selected for testing and the rest is used as a training dataset (LODO). For CRC classification, 4 different classification algorithms (Random Forest (RF), Logitboost, Adaboost, and Decision Tree (DT)) are used. The best performance among these methods was obtained with the Random Forest algorithm with an AUC of 0.98 by using JP for the training data set and JPN populations for the test data set in the cross-population performance evaluation. © 2023 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 3Protein İkincil Yapı Tahmini Için Makine Öǧrenmesi Yöntemlerinin Karşılaştırılması(Institute of Electrical and Electronics Engineers Inc., 2018-05) Aydin, Zafer; Kaynar, Oǧuz; Görmez, Yasin; Işik, Yunus EmreThree-dimensional structure prediction is one of the important problems in bioinformatics and theoretical chemistry. One of the most important steps in the three-dimensional structure prediction is the estimation of secondary structure. Due to rapidly growing databases and recent feature extraction methods datasets used for predicting secondary structure can potentially contain a large number of samples and dimensions. For this reason, it is important to use algorithms that are fast and accurate. In this study, various classification algorithms have been optimized for the second phase of a two-stage classifier on EVAset benchmark both in the original input space and in the space reduced using the information gain metric. The most accurate classifier is obtained as the support vector machine while the extreme learning machine is significantly faster in model training. © 2018 Elsevier B.V., All rights reserved.