Weather forecasting has been a major challenge due to the uncertain nature of the weather. Numerical models, such as the “Action de Recherche Petite Echelle Grande Echelle” (ARPEGE), the “Global Forecasting System”, the “European Center for Medium-Range Weather Forecasts”, are widely adopted by many meteorological services to forecast weather parameters. Under certain conditions, numerical models may have lower forecast accuracy, which is due to several factors such as the chaotic nature of the partial differential equations that simulate the evolution of the atmosphere and the difficulty of forecasting over countries with a fairly small surface area but with a very varied relief. This paper proposes a time series analysis approach based on the vector autoregression model (VAR) as an alternative and robust solution. The results are very promising (average of about 96.67% of precision between real values and three predicted parameters: temperature minimum, maximum humidity and precipitation) in the field of short-term weather parameter forecasting. In addition, the use of VAR models has solved the major problem posed by the chaotic equations of the ARPEGE model with greater accuracy on the one hand, and the execution time, forecasting accuracy and robustness of the SARIMA univariate models on the other.