Prediksi Evaporasi Berbasis Mesin: Perbandingan ANN, KNN, Random Forest dan Regresi Linier

Abstract

Evaporation plays a key role in water allocation, yet data limitations are often encountered. This study evaluates four regression models (Linear Regression, K-Nearest Neighbors, Random Forest, and Artificial Neural Network—ANN) to predict evaporation rates at the Banten Climatology Station. Models were assessed using R-squared (R²) and Root Mean Squared Error (RMSE). The results show that the ANN achieved the best accuracy with RMSE = 0.122 and R² = 0.475 (47.5%), followed by Linear Regression (RMSE = 0.123, R² = 0.460), K-Nearest Neighbors (RMSE = 0.126, R² = 0.437), and Random Forest (RMSE = 0.129, R² = 0.406). Other models also provided acceptable predictions, but the ANN stood out as the most accurate and reliable for applications at the Banten Climatology Station. These findings offer valuable insights for water resources management and agricultural planning, highlighting the potential of machine learning techniques to overcome evaporation data limitations.