Fuzzy time series has been developed for forecasting in many areas of real life applications. This paper presents an application of fuzzy time series in Malaysian Government direct tax collection. The historical data of Malaysian Government direct tax collection are tested using the method proposed by Singh to obtain forecasted values of direct tax. The forecasted values and actual values are compared to check its accuracy. The results indicate that the method is viable in direct tax forecasting with its mean square error is less than ten percent.