Yazar "Olmez, Emre" seçeneğine göre listele
Listeleniyor 1 - 5 / 5
Sayfa Başına Sonuç
Sıralama seçenekleri
Öğe A Decision Support System for Machine Learning-Based Determination of Zinc Deficiency: A Study in Adolescent Patients(Brieflands, 2024) Orbatu, Dilek; Bulgan, Zeynep Izem Peker; Olmez, Emre; Er, OrhanBackground: Over the past three years, zinc deficiency among adolescents has varied based on region and access to healthcare. Globally, zinc deficiency affects approximately 2 billion people, leading to serious issues such as immune problems and growth delays, particularly in developing countries. In the U.S., around 10% of adolescents experienced zinc deficiency in 2021, with a higher prevalence among teenage girls. In Europe, deficiency rates are generally low but can be significant in Eastern Europe and Central Asia. In Asia, particularly in rural and low-income areas, deficiency rates range from 20- 30%. In Turkey, the prevalence is high due to poor nutrition. Objectives: This study aimed to develop a machine learning-based decision support system to determine zinc deficiency in children and adolescents aged ID- 18 years. Methods: This machine learning-based study was conducted with 370 adolescents aged 10-18 years to assess their zinc deficiency. The dataset consists of 8 feature vectors and an output vector. The machine learning methods used in the analysis include logistic regression, naive bayes, decision tree (CART), K-nearest neighbors (K-NN), support vector machine (SVM), gradient boosting classifier, AdaBoost classifier; bagging classifier; random forest classifier; multilayer perceptron (MLP) classifier; and XGBoost (XGB) classifier. Evaluation metrics such as accuracy, precision, recall, and Fl score were used to assess the performance of these methods. Including specific values for these metrics, such as SVM achieved 94.6% accuracy, would allow readers to quicldy compare the effectiveness of the models. Different metrics serve various purposes: Accuracy provides an overall view of performance, precision and recall highlight specific aspects, and the Fl score balances precision and recall. Results: The mean age of the patients in the dataset was 13.79 +/- 1.18 years. Of the children, 6432% (n = 238) were female and 35.68% (n =B2) were male. It was found that 62.7% (n = 232) of the children had low zinc levels, while 373% (n = ox ) did not require zinc supplementation. Thirteen different machine learning methods were applied to a 70% training and 30% testing set. As a result, the SVM method provided the most successful outcome with 94.6% accuracy. Implementing the SVM-based system in pediatric clinics could improve efficiency and patient care by automatically detecting high-risk zinc deficiency patients based on lab results, providing early intervention alerts for faster treatment, and improving health outcomes. Highlighting these practical applications could increase the study's appeal to healthcare professionals by demonstrating its real-world benefits. Providing detailed information on these applications would enhance the study's clarity and practical value, making it more valuable for researchers and healthcare providers interested in Al tools for adolescent health. Conclusions: This study concluded that machine learning methods can effectively determine zinc deficiency in children. The SVM method demonstrated superior classification performance compared to the other methods. An SVM-based decision support system could be integrated into pediatric outpatient clinics to enhance diagnostic accuracy and patient care.Öğe Customer Churn Prediction Using Ordinary Artificial Neural Network and Convolutional Neural Network Algorithms: A Comparative Performance Assessment(Gazi Univ, 2023) Seymen, Omer Faruk; Olmez, Emre; Dogan, Onur; Orhan, E. R.; Hiziroglu, AbdulkadirChurn studies have been used for many years to increase profitability as well as to make customer -company relations sustainable. Ordinary artificial neural network (ANN) and convolution neural network (CNN) are widely used in churn analysis due to their ability to process large amounts of customer data. In this study, an ANN and a CNN model are proposed to predict whether customers in the retail industry will churn in the future. The models we proposed were compared with many machine learning methods that are frequently used in churn prediction studies. The results of the models were compared via accuracy classification tools, which are precision, recall, and AUC. The study results showed that the proposed deep learning-based churn prediction model has a better classification performance. The CNN model produced a 97.62% of accuracy rate which resulted in a better classification and prediction success than other compared models.Öğe EEG Based Cigarette Addiction Detection with Deep Learning(Int Information & Engineering Technology Assoc, 2024) Cay, Talip; Olmez, Emre; Altin, Cemil; Tanik, NerminIn this study, cigarette addiction detection was performed using machine learning techniques with time -frequency feature extraction methods on EEG data collected from 30 different male individuals. Electroencephalography (EEG) data collected from individuals who underwent the Fagerstr & ouml;m Test for Nicotine Dependence (FTND) were labeled as dependent or non-dependent based on their test results. The obtained EEG data were first subjected to Discrete Wavelet Transform (DWT). Then, Power Spectral Density (PSD) analysis and feature extraction processes were performed separately on the outputs obtained from the DWT process. The data obtained from PSD analysis and feature extraction processes were classified using Artificial Neural Networks (ANN). The aim of this study is to achieve higher success rates in cigarette addiction detection by classifying EEG data with machine learning methods after extracting time -frequency features, rather than using traditional methods. In this study, responses to cigarette stimuli were classified using machine learning methods based on EEG graphs. The results revealed that temporal and prefrontal lobes were more distinctive in responses to cigarette stimuli, and success rates were higher in the theta frequency band.Öğe EEG-Driven Biometric Authentication for Investigation of Fourier Synchrosqueezed Transform-ICA Robust Framework(Springer Heidelberg, 2023) Gorur, Kutlucan; Olmez, Emre; Ozer, Zeynep; Cetin, OnursalBiometric authentication systems have recently been gaining increased attention as an integral part of modern civic life. Security surveillance systems encompass broad categories ranging from unlocking mobile phones to personal identification authentication. Most state-of-the-art biometric systems employ fingerprint, face recognition, iris scanner systems, etc. These biometric systems are popular because of their easy-to-use and highly accurate in-person authentication. However, they do not guarantee liveness detection and can be easily deceived. Electroencephalography (EEG)-based biometric systems have dynamic and nonstationary characteristics during liveness, which offer unique, universal, and robust approaches against fraud attacks and thus present a high potential for secure biometric authentication systems. This study aimed to investigate the performance of a new framework for an FSST-ICA-based EEG-biometric authentication approach over motor imagery (MI) signals using an ensemble of LSTM deep models. The Fourier synchrosqueezed transform (FSST) was performed to implement feature extraction by analyzing the time-frequency (TF) matrix properties of the EEG signals. Synchrosqueezing transform was adopted as a feasible way to provide compact component localization capabilities for dynamic and nonstationary EEG signals with detailed spectral properties in the TF domain. Independent component analysis (ICA) was also carried out to decompose EEG multichannel sources in order to improve the true acceptance rate (TAR) and false acceptance rate (FAR) performance, as well as the correct recognition rate. The biometric authentication outcomes indicated that a high average accuracy (99.54%), sensitivity (99.81%), and specificity (99.41%) had been obtained regarding the one-versus-others discrimination among seven individuals via MI-EEG raw and subband (< 30 Hz) signals. Furthermore, high average TAR (97.8%) and low FAR (0%) values demonstrated robustness against multiple trials. To the best of our knowledge, an FSST-ICA framework for the EEG-based biometric approach using an ensemble of LSTM deep models has not been explored to date in the current literature. The study presents a highly secure and low-cost biometric system having broad fields of application. The effectiveness of the proposed framework over the spatiotemporal dynamics of the MI-EEGs was also evaluated by examining the broad statistical methods. This appears to be the first attempt to validate the discrimination of individuals in both raw and time-frequency features using extensive statistical analysis.Öğe STATISTICAL TECHNIQUES VS. MACHINE LEARNING MODELS: A COMPARATIVE ANALYSIS FOR EXCHANGE RATE FORECASTING IN FRAGILE FIVE COUNTRIES(Editura Ase, 2023) Bakir, Muhammed Rasid; Bakirtas, Ibrahim; Olmez, EmreIn 2013, the Federal Reserve (Fed) announced the end of its expansionary monetary policy, which had a significant impact on certain countries. These countries, colloquially referred to as the fragile five, were heavily dependent on financial capital flows, which led to deviations from inflation targets due to the exchange rate pass-through effect. Consequently, monetary authorities and other financial actors need accurate exchange rate forecasts to mitigate these deviations and improve the effectiveness of monetary policy. This study aims to forecast the exchange rates of the fragile five countries using both traditional statistical methods and machine learning techniques. The traditional statistical methods used in this study include Naive Drift, Theta, Holt's Exponential Smoothing and ARIMA models, while the machine learning methods include RNN, LSTM, GRU and CNN architectures. The results show that machine learning methods outperform traditional statistical methods in terms of prediction accuracy for all countries. While statistical methods show a directional accuracy rate between 47% and 60%, RNN, one of the machine learning models, shows an accuracy rate between 80% and 90%. Overall, these results suggest that machine learning methods can provide more accurate exchange rate forecasts for the fragile five countries than traditional statistical methods. These findings may be valuable for monetary authorities and financial actors seeking to improve the effectiveness of monetary policy in these countries.
