Performance comparison of LS-SVM and ELM-based models for precipitation prediction in Barak valley: A case study

In weather forecasts, precipitation is an essential topic. Prediction of precipitation poses a difficult task since it depends on several parameters like minimum and maximum temperature, wind speed, humidity that change timely. Also, weather calculation fluctuates with atmospheric variables and geographical location. Hence, various artificial intelligence (AI) methods have been utilized for understanding importance of each predictor and the predicted (precipitation). The present study validates a computationally fast, simple, and effective non-linear algorithm called extreme learning machine (ELM) to predict precipitation considering data from two rain gauge stations located in Barak valley, Assam, India. For demonstrating efficacy of the proposed ELM algorithm, a performance assessment was conducted in terms of prediction abilities with Least Squares-Support Vector machine (LS-SVM). Accurateness of ELM is evaluated utilising statistical measures of Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (R2). It was observed that ELM showed promising outcomes with RMSE-40.29, MAE-19.36, and R2-0.97412 at Karimganj station during the testing period. Also, results showed better generalisation capability besides faster computation time. Essentially, applied techniques and obtained outcomes will improve our understanding and enable future researches in regards to long-term precipitation prediction in other areas.