Machine Learning Meets with Monsoon: Improving Rainfall Predictions for Agriculture and Tourism in North Bengal

Rainfall patterns significantly impact agriculture and tourism, especially in regions like North Bengal, India, where the local economy heavily depends on both sectors. Landslides due to heavy rainfall create a severe threat to tourism in the districts of Darjeeling and Kalimpong. As far as we know, this is the first time a rainfall prediction study has been done in the Darjeeling and Kalimpong districts of North Bengal, India, using different machine-learning algorithms. The analysis uses a comprehensive dataset for 43 years with 32 weather parameters, including all-sky, UV, temperature, humidity, wind speed, and precipitation data. Feature selection uses SelectKBest with F-regression to identify the most relevant variables. The study estimates the performance of multiple machine learning algorithms, such as Ad-aboost Regression, Decision Tree Regression, Support Vector Regression, XGBoost Regression, LightGBM Regression, Random Forest Regression, and KNN Regression. The results indicate that Random Forest Regression exhibit the highest accuracy in predicting rainfall for these regions, helping local agricultural and tourism planning.