Araştırma Çıktıları | WoS | Scopus | TR-Dizin | PubMed
Permanent URI for this communityhttps://hdl.handle.net/20.500.14719/1741
Browse
5 results
Search Results
Publication Metadata only Performance Comparison of Oral, Laryngeal and Thoracic Sounds in the Detection of COVID-19 by Employing Machine Learning Techniques(IEEE, 2022) Gozuacik, Necip; Serbes, Gorkem; Kara, Eyup; Atar, Eren; Sakar, C. Okan; Yener, H. Murat; Borekci, Sermin; Korkmazer, Bora; Karaali, Ridvan; Kara, Halide; Gulmez, Zuleyha; Cogen, Talha; Atas, Ahmet; Bahcesehir University; Yildiz Technical University; Istanbul University; Istanbul University - CerrahpasaCOVID-19 can directly or indirectly cause lung involvements by crossing the upper airways. It is essential to quickly detect the lung involvement condition and to follow up and treat these patients by early hospitalization. In recent COVID-19 diagnosis procedure, PCR testing is applied to the samples taken from the patients and a quarantine period is applied to the patient until the test results are received. As a complement to PCR tests and for faster diagnosis, thin-section lung computed tomography (CT) imaging is used in COVID-19 patients. In this study, it is aimed to develop a method that is as reliable as CT, and compared to CT, less risky, more accessible, and less costly for the diagnosis of COVID-19 disease. For this purpose, first speech and cough sounds from the oral, laryngeal and thoracic regions of COVID-19 patients and healthy individuals were obtained with the multi-channel voice recording system we proposed, the obtained data were processed with machine learning methods and their accuracies in COVID-19 diagnosis were presented comparatively. In our study, the best results were obtained with the features extracted from the cough sounds taken from the oral region.Publication Metadata only Natural Language Processing-Based Product Category Classification Model for E-Commerce(IEEE, 2022) Koksal, Deniz; Alacan, M. Mert; Olgun, Ecesu; Sakar, C. Okan; Bahcesehir UniversityIn scope of the great dynamism that the world of e-commerce gained with the post-pandemic changes on customer behavior, extending the customer's time on site has become much more valuable. With regards to creating a better understanding of the online shopper's intent on site, an effective search engine is the best tool for improving the user experience. A useful search engine needs to produce fast results and do intent analysis of customers correctly to present successful recommendations. In this study, BERT, ELECTRA and RoBERTa which are language models that give successful results in similar problems and pre-training libraries in Turkish languages are used for developing text classification and customer intent analysis models based on e-commerce product categories. The results of these deep learning models, which were optimized using the product description and comment libraries of the selected e-commerce brand, are shared in a comparative way, and the results of our e-commerce end-user intent analysis model, which was tested on user search history, were detailed on the basis of product category. The developed model can be used for e-commerce product prioritization in Turkish language, as well as creating an infrastructure for the development of intent analysis models that are able to predict as far as the main product that the user wants to display in searches containing multiple products.Publication Metadata only A Dynamic Recurrent Neural Networks-Based Recommendation System for Banking Customers(IEEE, 2021) Avci, Hasan; Sakar, C. Okan; Bahcesehir UniversityIn recent years, machine learning approaches are replacing traditional methods in many industries. In parallel with these developments, marketing operations in banking started to be supported with machine learning based applications. In this study, a highly applicable product recommendation approach for banking customers was implemented using deep learning techniques. For this purpose, the dynamic recurrent neural network (DREAM) architecture, which was previously applied to e-commerce data, has been applied to banking customer data for recommendation system design. Comparative experiments with long short term memory and multilayer perceptron-based solutions have shown that the DREAM based recommendation approach has significant potential to be used in product proposal in the banking sector.Publication Metadata only Hand-Gun Detection in Images with Transfer Learning-Based Convolutional Neural Networks(IEEE, 2020) Veranyurt, Ozan; Sakar, C. Okan; Bahcesehir UniversityThis study aims to evaluate the performance of different deep learning methods based on CNN (Convolutional Neural Networks) for hand-gun detection from images. Within the context of object detection, which is an application of CNN, this study further focusses on hand-gun detection from images. The application of a CNN trained from scratch, transfer learning method using VGG-16 (Visual Geometry Group) model and fine-tuning based on the same model are applied on the same image dataset and the results are evaluated using various evaluation metrics. The results on 8300 images with and without a hand-gun showed that the highest accuracies are obtained with fine tuning method. Besides, it has been observed that the CNN model obtained with fine-tuning method gave balanced accuracies on the images containing and not containing hand-gun.Publication Metadata only Comparison of Transformer-Based Models Trained in Turkish and Different Languages on Turkish Natural Language Processing Problems(IEEE, 2022) Aytan, Burak; Sakar, C. Okan; Bahcesehir UniversityTransformer-based pre-trained language models have yielded successful results in natural language processing (NLP) problems in recent years. In these approaches, the models are trained in an unsupervised manner using a large corpus, with the use of mechanisms such as masking and next sentence prediction. In sub-NLP problems, these models are fully or partially updated with a fine-tuning approach, or the vectors obtained from these models are directly mapped to the output with neural network layers added. In this study, transformer-based BERT, RoBERTa, ConvBERT and Electra models, which have given successful results for different languages and problems in the literature, have been applied to Turkish sentiment analysis, text classification and named entity recognition problems, and the results are presented comparatively. One of the contributions of the study is to train the RoBERTa model on a total of 38 GB Turkish corpus to be shared as open source for use in Turkish problems in the field of NLP. Experimental results have shown that the RoBERTaTurk model gives comparable results to the other transformer-based models in Turkish NLP problems. In addition, language models trained in different languages were also tested on Turkish classification problems. According to the obtained results, as the size of the labeled training set increases, the success of the language models trained in different languages becomes close to the success of the language models trained for the Turkish language. It has also been shown that Turkish language models give better results when the number of labeled data is low.