https://hdl.handle.net/20.500.12960/30 2026-05-10T15:35:39Z https://hdl.handle.net/20.500.12960/1816 Prediction of traffic accidents trend with learning methods: a case study for Batman, Turkey Bakış, Enes; Erçetin, Mehmet Ali; Acar, Emrullah; Gökalp, İslam; Yılmaz, Musa Assessing the trend of fatalities in recent years and forecasting road accidents enables society to make appropriate planning for prevention and control. This study analyses the road traffic accident data between the years 2013 and 2022 obtained for the province of Batman in Turkey, where it has not been considered before. The scope of the data analysed includes the fatalities and injuries of drivers, passengers and pedestrians. The road accident forecast for the next ten years up to 2032 is the focus of this study and numerous analyses using learning methods such as State Space Models (SSM), Artificial Neural Networks (ANN), Autoregressive Integrated Moving Average (ARIMA) and hybrid models (CNN + LSTM and Attention + GRU) have been performed on the available data. The predictions made with the above models give results with acceptable accuracy. However, they give different results depending on the parameters used. The models created with the data studied show that the number of road accidents and the related deaths and injuries will continue to increase over the next 10 years, starting in 2022. If the causes of road accidents are not eliminated and the situation remains stable as it is in 2022, the number of accidents, deaths and injuries is expected to double by 2032. 2025-01-01T00:00:00Z https://hdl.handle.net/20.500.12960/1811 Comparative analysis of shallow and hybrid deep learning models for predicting the cooling efficiency of nanofluid-cooled photovoltaic panel across multiple materials Özdemir, Yunus; Budak Ziyadanoğulları, Neşe; Bakış, Enes; Acar, Emrullah Active cooling can mitigate temperature-induced performance losses in photovoltaic (PV) modules, and nanofluids are a promising coolant option. This study develops data-driven models to predict the cooling efficiency of an actively cooled PV panel using seven working fluids: water and Al₂O₃/TiO₂ nanofluids at 0.01%, 0.1%, and 1 vol%. For each fluid, outdoor measurements were collected over six hours at 30-min intervals (13 observations), including inlet/outlet temperatures, electrical variables of cooled and reference panels, and ambient conditions. Shallow regression models (Bayesian Ridge, SVR-RBF, Random Forest) were evaluated using leave-one-out cross-validation, and a hybrid deep learning (CNN+LSTM) model was also tested using k-fold cross-validation. Bayesian Ridge achieved the most consistent performance across materials (RMSE ≈ 0.35–0.66; R² ≈ 0.88–0.98). The hybrid CNN+LSTM reached RMSE as low as 0.28 with R² up to 0.98. SHAP-based interpretability analysis indicates that ambient temperature, irradiance, and the cooled-panel electrical variables are among the most influential predictors. These results show that lightweight ML models can reliably estimate PV cooling performance and reduce repetitive experimentation. 2026-01-01T00:00:00Z https://hdl.handle.net/20.500.12960/1773 Deep Learning-Based Time Series Prediction of Micro Gas Turbine Power Output Bakış, Enes; Acar, Emrullah Micro gas turbines play an important role in energy production, and it is critical to make accurate predictions in order to ensure that these turbines operate efficiently. The objective of this study is to predict the power output of micro gas turbines using three different deep learning models (artificial neural network (ANN), convolutional neural network (CNN) and long short-term memory (LSTM)) based on the input voltage data. A time series approach is employed to model the impact of the input voltage on power generation. Another objective of the study is to analyze the efficacy of deep learning-based methods for predicting the power output of micro gas turbines. The results indicate that the ANN model has a MAPE value of approximately 16%, the CNN model has a value of about 26%, and the LSTM model has a value of almost 20%. The ANN model demonstrates the lowest error rate, while the LSTM and CNN models exhibit satisfactory performance. These findings indicate that time series data can be effectively used to predict the power output of micro gas turbines. It is anticipated that in the future, the approaches will be further developed and applied more widely in industry as larger data sets and more sophisticated model improvements become available. 2024-01-01T00:00:00Z https://hdl.handle.net/20.500.12960/1772 Prediction of Electricity Production from Wind and Solar Energy by Employing Regression Models Örenç, Sedat; Acar, Emrullah; Özerdem, Mehmet Sıraç; Bakış, Enes Accurately predicting electricity production from renewable energy sources is critical for sustainable energy systems. Wind and solar energy play key roles in reducing greenhouse gas emissions and reliance on fossil fuels. However, forecasting their production is challenging due to weather variability and seasonal changes. This study employs advanced regression models to predict energy production, including Linear Regression, Gradient Boosting, AdaBoost, XGBoost, and Random Forest. The dataset comprises hourly wind and solar energy data from France in 2020, containing 59,807 samples. Model performance was evaluated using Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R2 metrics. Random Forest demonstrated the best performance, with an MAE of 219.30 MW, RMSE of 391.11 MW, and R2 of 0.97. XGBoost also showed strong results, achieving an MAE of 729.43 MW, RMSE of 1200.45 MW, and R2 of 0.75. These results highlight the Random Forest model's superior accuracy and reliability in capturing complex patterns. This work contributes to improving the integration of renewable energy into power grids by enhancing forecasting precision. 2024-01-01T00:00:00Z