https://hdl.handle.net/20.500.12294/422 Bilgisayar Mühendisliği Bölümüne ait makaleler bu koleksiyonda listelenir. Tue, 10 Nov 2026 23:31:13 GMT 2026-11-10T23:31:13Z https://hdl.handle.net/20.500.12294/4066 AutoFusion of feature pruning for decision making in operations research Atas, Pinar Karadayi; Akyuz, Sureyya Ozogur Recently, the fusion of algorithms in machine learning studies has taken a lot of attention, emphasizing the power of communal decision-making over-relying on a single decision-maker. One of the crucial questions in the aggregation of algorithms is which and how many models should be combined to achieve both the best accuracy and low complexity. It is already known in machine learning that as the complexity of the model increases too much, prediction accuracy decreases. There is a trade-off between these two features. In order to answer such questions, the diversity notion gets involved in overall consensus models. It is also shown that diversity alone does not determine the best ensemble (fusion), so accuracy and diversity together have been taken into account recently in such problems. We took into account those two notions simultaneously so that the number of algorithms and which algorithms should be in the ensemble is answered while solving the feature selection problems. The proposed method in this work is unique in that it includes an optimization model in the pruning phase, which finds the cardinality of the ensemble optimally. Using this optimization model, the size of the ensemble is found directly from the optimization model, instead of considered as a hyper-parameter. Our study shows a significant improvement in accuracy that achieves 0.702 on average among 8 datasets when compared to an unpruned case of 0.625. These results highlight the efficiency of our method both in model accuracy and in obtaining an optimal model complexity. We have validated our algorithm on different domains of data sets which shows better prediction accuracy values than existing ensemble-based feature selection methods. © 2024, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature. Mon, 01 Jan 2024 00:00:00 GMT https://hdl.handle.net/20.500.12294/4066 2024-01-01T00:00:00Z https://hdl.handle.net/20.500.12294/4060 Exploring the Molecular Interaction of PCOS and Endometrial Carcinoma through Novel Hyperparameter-Optimized Ensemble Clustering Approaches Atas, Pinar Karadayi Polycystic ovary syndrome (PCOS) and endometrial carcinoma (EC) are gynecological conditions that have attracted significant attention due to the higher prevalence of EC in patients with PCOS. Even with this proven association, little is known about the complex molecular pathways that connect PCOS to an increased risk of EC. In order to address this, our study presents two main innovations. To provide a solid basis for our analysis, we have first created a dataset of genes linked to EC and PCOS. Second, we start by building fixed-size ensembles, and then we refine the configuration of a single clustering algorithm within the ensemble at each step of the hyperparameter optimization process. This optimization evaluates the potential performance of the ensemble as a whole, taking into consideration the interactions between each algorithm. All the models in the ensemble are individually optimized with the suitable hyperparameter optimization method, which allows us to tailor the strategy to the model's needs. Our approach aims to improve the ensemble's performance, significantly enhancing the accuracy and robustness of clustering outcomes. Through this approach, we aim to enhance our understanding of PCOS and EC, potentially leading to diagnostic and treatment breakthroughs. Mon, 01 Jan 2024 00:00:00 GMT https://hdl.handle.net/20.500.12294/4060 2024-01-01T00:00:00Z https://hdl.handle.net/20.500.12294/4034 Early Software Defects Density Prediction: Training the International Software Benchmarking Cross Projects Data Using Supervised Learning Tahir, Touseef; Gencel, Cigdem; Rasool, Ghulam; Umer, Tariq; Rasheed, Jawad; Yeo, Sook Fern; Cevik, Taner Recent reviews of the literature indicate the need for empirical studies on cross-project defect prediction (CPDP) that would allow aggregation of the evidence and improve predictive performance. Most empirical studies predict defects at granularity levels of method, class, file, and module/package during the coding phase, and thereby avoid external failure costs. The main goal of this study is to perform an empirical study on early defect prediction at the beginning of a project at the product level of granularity for using it as input in planning quality activities of the project. Hence, both internal and external failure costs could be avoided as much as possible through proper planning of quality. We first made a systematic mapping study (SMS) on secondary studies (literature reviews) on defect prediction to identify the most used datasets, the project attributes and metrics utilized as estimators, and the supervised learning methods employed for training the data. Then, we made an empirical study on defect density prediction using cross-project data. We collected 760 project data from the International Software Benchmarking (ISBSG) dataset version 11, which reported both defects and functional size attributes. We trained the prediction models using: i) the complete set of project attributes, ii) the individual attributes, and iii) multiple subsets of attributes. We employed classification and regression approaches of machine learning. The machine learning models are trained using original values of the dataset, and z-score and logged transformations of original values to explore the effects of data normalization on prediction. Most machine learning models trained on the z-score transformation of the dataset performed best for classifying defects. The Multilayer-Perceptron (Neural Network) model trained on the z-score transformation of complete dataset predicted defects with the highest F1-score of 0.89 using binary classification. The logged transformation and feature selection methods improved the results for multivariable regression. The multivariable regression predicted defects with the highest Root Mean Squared Error (RMSE) and R2 (r-squared) values of 0.4 and 0.9, respectively, with a subset of 11 features using logged transformation. The results of classification and regression approaches indicate that defects can be predicted with reasonable accuracy at the software product level using cross-project data. © 2013 IEEE. Sun, 01 Jan 2023 00:00:00 GMT https://hdl.handle.net/20.500.12294/4034 2023-01-01T00:00:00Z https://hdl.handle.net/20.500.12294/4023 Reversible Logic-Based Hexel Value Differencing-A Spatial Domain Steganography Method for Hexagonal Image Processing Cevik, Taner; Cevik, Nazife; Rasheed, Jawad; Asuroglu, Tunc; Alsubai, Shtwai; Turan, Mehmet The field of steganography has witnessed considerable advancements in square-pixel-based image processing (SIP). However, the application of steganography in Hexel (Hexagonal pixel)-based Image Processing (HIP) is still underexplored. This study introduces a pioneering spatial steganography method called the Reversible Logic-Based Hexel Value Differencing (RLBHVD) method in the HIP domain. Our approach draws inspiration from Pixel-Value-Differencing (PVD), a SIP fundamental spatial-domain (S-D) steganography method. Initially, the image is transformed into the HIP domain using the custom software infrastructure developed for this project. Due to the absence of commercial equipment capable of producing HIP-domain images, traditional digital imaging systems are employed with their sensor components, analog-to-digital conversion units, and square-pixel-based displays. Once the image is converted, it is partitioned into standardized heptads, each comprising seven hexels. Simultaneously, the secret message is segmented for embedding into the hexels within each heptad. Unlike SIP-domain PVD, which embeds segments into independent pixel pairs, our method performs iterative embedding within each heptad. Additionally, we leverage Feynman gates, a core element of reversible logic, to achieve retrieval of both the cover image and the secret message. Unlike PVD in SIP, our approach enables reversibility in the recovery process. Experimental results demonstrate that our proposed method, RLBHVD, outperforms its SIP counterpart, PVD, by achieving a low Mean Squared Error (MSE), high Peak Signal-to-Noise Ratio (PSNR), and significant similarity between the stego-image and cover image histograms. These findings highlight the efficacy and superiority of our HIP-based steganography approach in comparison to existing SIP methods. Sun, 01 Jan 2023 00:00:00 GMT https://hdl.handle.net/20.500.12294/4023 2023-01-01T00:00:00Z