SPEI Based Drought Prediction and Validation Using Machine Learning Algorithm in Bangladesh from 1981 to 2017

A natural event known as a drought is defined by a long period of insufficient rainfall and limited access to water supplies, which means there isn't enough water for many uses, including agriculture, industry and domestic consumption. The Standardized Precipitation Evapotranspiration Index (SPEI) is a frequently utilized indicator in drought monitoring. In this work, from 1981 to 2017, temporal variation at Ishwardi upazila was used to evaluate and analyze the SPEI's performance at 1, 3, 6 and 12-month timeframes. An analysis is carried out to see how well machine learning approach such as Support Vector Machine (SVM), Random-Forest (RF), Decision-Tree (DT), Extreme Gradient Boosting (XGB) and K-Nearest Neighbors (KNN) work in forecasting drought conditions. The prediction outcomes of all five techniques were assessed using four performance measures: the Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and Nash-Sutcliffe Model Efficiency Coefficient (NSE). The proposed methods provided good results and can serve as a quick decision-making tool for water management.