Invited Speakers of ICOMS 2026

I conduct academic research on artificial neural networks, artificial intelligence optimization algorithms, fuzzy inference systems, and time series forecasting. I have published over 80 articles and conference papers in these fields. Since 2025, I have served as a professor in the Department of Data Science and Analytics at Giresun University. I served as a visiting researcher at Brunel University London in the UK. I have over 15 years of professional experience in academia. I have teaching experience covering a wide range of artificial intelligence courses at both the undergraduate and graduate levels. I received my Ph.D. in 2014, my master’s degree in 2011, and my bachelor’s degree in 2009 from Ondokuz Mayıs University. In the “Last 5 Years – Artificial Neural Network” ranking published by ScholarGPS, I ranked 22nd globally in 2024 and 10th in 2025. Additionally, I am listed in the “Top 2% Most Influential Scientists in the World” list published by Stanford University in the “Annual Impact” category for both 2024 and 2025.

Speech Title: "Decile Mean-Based Artificial Neural Network"

Abstract: Multilayer perceptron (MLP) models have been widely used for time series forecasting; however, their performance degrades significantly in the presence of outliers due to the sensitivity of mean-based aggregation mechanisms. To address this limitation, robust neural architectures based on alternative aggregation functions have been proposed, though many rely on median-based structures or remain sensitive to certain contamination patterns. In this study, a novel model, namely Decile Mean Artificial Neural Network (DM-ANN), is proposed to enhance robustness against outliers while preserving statistical efficiency. Unlike conventional neural networks, DM-ANN employs a decile mean-based aggregation function, providing a balance between the robustness of the median and the efficiency of the mean. The proposed architecture is trained using the Artificial Bee Colony (ABC) algorithm. The model is evaluated on financial time series datasets under both clean and contaminated scenarios. Experimental results show that DM-ANN achieves superior forecasting accuracy and exhibits strong robustness under varying levels of outlier contamination.

.

Bio: Özlem Kaymaz received the Ph.D. degree in Biostatistics from Ankara University, Ankara, Türkiye, in 2015. She conducted postdoctoral research at the University of Leeds, Leeds, U.K. She is currently an Associate Professor in the Department of Statistics, Ankara University, Ankara, Türkiye. Her research interests include biostatistics, statistical modeling, multivariate statistical analysis, and statistical inference. She has published numerous articles in international peer-reviewed journals and serves as a reviewer for SCI-indexed journals in statistics, data science, and interdisciplinary research. She serves as the principal investigator and researcher in several nationally funded interdisciplinary research projects. Her current research aims to develop innovative statistical and machine learning approaches for biomedical, environmental, and public health applications.

Speech Title: "Predicting Air Quality Index Using Robust Machine Learning Methods"

Abstract: Air pollution is recognized as a significant risk factor for both the environment and public health. While air pollutants directly affect air quality, meteorological factors have an indirect effect. Observations of air pollutants and meteorological factors often contain outliers. These outliers may arise from technical causes such as instrument malfunctions, calibration problems and incorrect data entry as well as from sudden meteorological changes and unexpected industrial emissions. Outliers can significantly affect the accuracy and performance of predictive models. This study aims to predict the Air Quality Index (AQI) by integrating robust loss functions (Huber, Tukey and Hampel) into machine learning algorithms such as support vector regression (SVR), random forest (RF) and extreme gradient boosting (XGBoost). The objective is to identify the method that is most robust to outliers and provides the best predictive performance. The data consist of daily air pollution and meteorological measurements collected from Ankara Sıhhiye station between 2022 and 2025. Air pollutant variables include particulate matter (PM2.5, PM10), nitrogen dioxide (NO₂), nitric oxide (NO), nitrogen oxides (NOₓ), sulfur dioxide (SO₂), and ozone (O₃), while meteorological variables include temperature, humidity, wind speed, wind direction and pressure. The findings indicate that robust machine learning models produce more accurate and reliable AQI predictions in the presence of outliers. This study proposes a modeling strategy that is both statistically robust and highly accurate and contributing to the development of data-driven decision support systems for improving air quality management.

.