Rainfall Prediction Using Adaptive Neuro-Fuzzy Inference System Method

This study analyzes rainfall prediction using the Adaptive Neuro-Fuzzy Inference System (ANFIS) method to improve model accuracy, particularly in extreme rainfall events. The objective of this study is to evaluate rainfall prediction using the ANFIS method to enhance model accuracy, especially in predicting extreme rainfall occurrences. The results indicate a moderate positive correlation with R² of 0.55, demonstrating good model performance at low rainfall levels (<20 mm) but a tendency to underestimate high-intensity rainfall (>60 mm). Residual analysis reveals a distribution around zero without systematic bias, though significant outliers (>20 or <-20) suggest the need for accuracy improvement. Monthly RMSE exhibits fluctuations, with the best performance observed in the June-July-August (JJA) season and notable challenges in December-January-February (DJF) due to extreme variability. Annual RMSE is also higher in extreme rainfall years (2018, 2023) compared to stable years (2019, 2020). The implementation of ANFIS enhances prediction sensitivity by incorporating additional variables such as temperature and humidity, leading to more accurate forecasts, particularly in extreme weather conditions. This study is further supported by STEM research at Universitas Syiah Kuala, which emphasizes the importance of artificial intelligence in climate data analysis to improve weather prediction accuracy. The ANFIS-based approach applied in this research aligns with STEM studies, which highlight the integration of artificial intelligence in meteorology to mitigate hydrometeorological disaster risks