EVALUATION OF THE GRADIENT BOOSTING ALGORITHM BASED ON TRAIN/TEST RATIOS IN SOLAR ENERGY POWER GENERATION FORECASTING

Hits: 79
  • TITLE: EVALUATION OF THE GRADIENT BOOSTING ALGORITHM BASED ON TRAIN/TEST RATIOS IN SOLAR ENERGY POWER GENERATION FORECASTING
  • AUTHOR(S): D. Akal, T. Tez, İ. Umut.
  • ABSTRACT: Accurate forecasting of solar energy generation is of critical importance for energy planning, resource management, and sustainability efforts. This study investigates the performance of the Gradient Boosting algorithm in predicting solar power output. The analysis utilizes the Solar Energy Power Generation Dataset obtained from the Kaggle platform. The dataset comprises hourly meteorological variables such as temperature, humidity, pressure, precipitation, various cloud cover types, shortwave radiation, wind speed and direction, solar angles, as well as the corresponding power generation values. During the preprocessing phase, the data were imported into the Orange open-source data analysis software, where variable names were standardized and transformed into a format suitable for modeling. Gradient Boosting was selected as the predictive algorithm, and its performance was evaluated under various train/test split ratios (50%, 60%, 66.6%, 70%, 75%, 80%, 90%, and 95%). Several essential performance metrics including the coefficient of determination (R²), root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) were employed to assess the model's performance. The highest R² value (0.790) and the lowest error rates were achieved with a 90% training ratio (RMSE=428.959, MAE=289.195). However, a slight performance decline observed at the 95% training ratio suggests a potential risk of overfitting. Overall, the findings demonstrate that Gradient Boosting is a reliable and effective method for forecasting solar energy generation, with optimal results obtained at the 90% training level. Future studies may achieve higher accuracy and generalization capacity through the integration of alternative boosting algorithms and hyperparameter optimization techniques.
  • DOI: http://doi.org/10.62853/HHZK1143
  • PAGES: 43-46
  • DOWNLOAD: Vol71-2025-9-43-46.pdf
  • HOW TO CITE THIS ARTICLE: D. Akal, T. Tez, İ. Umut. Evaluation of the gradient boosting algorithm based on train/test ratios in solar energy power generation forecasting. Journal of the Technical university of Gabrovo. 71 (2025) 43-46.