Drought is among the most important natural disasters influencing different aspects of human life. In recent decades, intelligent techniques have shown to be highly capable of modeling and forecasting nonlinear and dynamic time series. Hence, the present study aimed to forecast drought using and comparing the multilayer perceptron artificial neural network (MLP ANN), adaptive neuro-fuzzy inference systems (ANFIS), support vector machine (SVM) model, and the autoregressive integrated moving average (ARIMAX) multivariate time series. To this end, the precipitation data obtained from the Yazd synoptic station for a 51-year statistic period were used. Moreover, the humidity levels for short-term (3 and 6 months) and long-term (9, 12, 18, and 24 months) periods were calculated using the Standardized Precipitation Index (SPI). Next, based on the results of calculations, the 1961–2002 period was selected as the control group and the 2003–2012 period was selected as the experimental group. In order to forecast the SPI for the t + 1 period, values of SPI, precipitation, and temperature of previous eras were used. Results indicated that in a 9-months period (as the timescale), the ARIMAX model gives SPI values and forecast drought with more precision than the SVM, ANFIS, and MLP models.