Drought is a major natural hazard that intensely affects all environmental parts. Drought forecasting with suitable accurateness remarkably aids drought management, and hence minimizes damages caused by drought. In recent years, several artificial intelligence approaches are utilized for drought forecasting. This study emphasizes on investigating and comparing forecasting abilities of artificial neural networks (ANN), support vector machine (SVM), wavelet-ANN (W-ANN), and wavelet-SVM (W-SVM) for drought forecasting in Bharuch, Porbandar, Surendranagar, India. Models utilized in present study are developed based on standard precipitation index (SPI). SPI was computed for 3, 6, 9, 12, and 24 month timescales on the basis of monthly rainfall data over a 35-year period from 1990 to 2019. Performance of all models are evaluated using different performance measures, namely Wilmton Index (WI), root mean squared error (RMSE), and mean squared error (MSE). An assessment amid actual data and predicted data demonstrates that WI values of W-SVM model were 0.98199, 0.98207, and 0.98231, for Bharuch, Porbandar, Surendranagar stations, respectively. It can be observed that wavelet transform can develop drought forecasting models. Also, it is observed that SVM models produced more precise prediction results than ANN models. With respect to model accurateness, performance of models is presented in the following respective order: W-SVM, W-ANN, SVM, and ANN. Finally, all proposed forecast models can be used for early drought warning systems.