This paper presents an alternative to forecasting natural gas consumption in Medellín (Antioquia), the second largest city in Colombia. The data for this study was collected from the SUI (Sistema Único de Información), which is the official database for gas consumption rates and prices in Colombia. Dataset covered the first quarter 2003 up to the second quarter 2014, using ARIMA and Support Vector Machine models predictions, forecast performance was estimated. Results indicate that the ARIMA model presents a better fit than the SVM model in terms of RSME and MAE.