Meteorological drought is an early warning system of drought conditions that are transferred to environmental drought and can seriously affect the aquatic ecosystem. Therefore, monitoring and forecasting drought and establishing an early warning and foresight system in drought-prone areas is inevitable. In the present study, the performance of ARIMA and SARIMA time series models was compared with ANFIS and ELM machine learning algorithms to predict meteorological drought in Khuzestan Province. In order to evaluate the SPI index at one, three, six and 12 month time scales, statistics and information from eight synoptic stations were used during the years 1989 to 2020. Then, time series and machine learning models were run based on SPI and finally the best model for drought prediction was extracted. In order to evaluate the performance of the models, the evaluation criteria of RMSE, MAE, NS and R were used. The results showed that machine learning algorithms had higher accuracy than time series models in predicting SPI, and among the models studied, the ELM model had the lowest dispersion, the highest accuracy, and the highest correspondence with the first and third quadrant bisectors. Also, based on the SPI results, it was determined that in the study area, drought is in a normal state, but a more severe situation could arise in the future. Therefore, using the results of various modeling in forecasting and warning systems can significantly help in managing and reducing the damages caused by this destructive environmental phenomenon.