Performance Evaluation of Nature-Inspired Optimization Algorithms vs. the Tweedy Model in Rainfall Data Forecasting: A Case Study in Kirkuk, Iraq

Accurate rainfall prediction is crucial for agricultural water management. However, complex data pose a challenge to traditional statistical models. Therefore, this study assesses the effectiveness of a series of predictive algorithms in analysing continuous data. We looked at traditional and contemporary algorithms and compared four major algorithms: Tweedie, WPO, XGBoost and HBOA. These models were evaluated with the help of statistical parameters such as coefficient of determination (R²), mean absolute error (MAE) and root mean square error (RMSE). According to the results, the WPO algorithm performed better in predicting the results. It showed the peak value for the coefficient of determination (R² = 0.684) and the minimal value for the root mean square error (RMSE = 10.58). Conversely, the Tweedie model demonstrated consistent and stable performance (R² = 0.596 and RMSE = 11.97). It constitutes an appropriate option for predictions that require a balance between accuracy and reliability. The HBOA algorithm attained ordinary performance metrics, characterised by a comparatively balanced statistical profile (R² = 0.621, MAE = 7.71, and RMSE = 11.59). However, the XGBoost algorithm performed worst, as evidenced by the lowest coefficients of determination and the highest error metrics (R² = 0.35, RMSE = 15.18, and MAE = 10.85). These results indicate a restricted efficacy when applied to this specific dataset. Accordingly, the WPO algorithm is recommended for predicting continuous data due to its highest performance. Nevertheless, the Tweedie model is a viable option when a balance between performance and accuracy is required.