Generalized Autoregressive Conditional Heteroskedasticity (GARCH) is one of the most popular time series models that can be used for time series forecasting. However, the computation of the long run variance in the GARCH model is based on the historical data that does not reflect the influence of the recent variance. This study proposed the sliding window GARCH (SWGARCH) model, which is an enhancement of the GARCH model to overcome the limitation of the variance. The sliding window technique is solely to estimate the variance in the SWGARCH model. A performance evaluation of SWGARCH was performed on Standard and Poor's 500 index dataset and compared with two (2) common time series forecasting models in terms of mean square error and mean absolute percentage error. The experimental results showed that the performance of SWGARCH is superior than GARCH and ARIMA-GARCH, which confirmed that SWGARCH can be used for time series forecasting. © 2017 Association for Computing Machinery.