A Comparative Study of LGBM-SVR Hybrid Machine Learning Model for Rainfall Prediction

Weather forecasting is a critical factor in determining the crop production and harvest of any geographical location. Among various other factors, rainfall is a crucial determining component in the sowing and harvesting of crops. The aim and intent of this paper is to analyze various machine learning algorithms like LightGBM and SVR, and develop a hybrid model using LightGBM and SVR to accurately predict rainfall. The hybrid model implements both LightGBM and SVR on a preprocessed dataset and then combines the predicted values of the results through an ensemble model which considers the average of these values based on a predefined weight. The weight of the model is determined by considering various combinations, and the result with the least error is taken into consideration for that particular dataset. The study shows that the hybrid model performed better than LightGBM and SVR individually, and produced the least root mean square error yielding a more accurate prediction of rainfall.