Forecasting process play an important role in time series data as required for decision-making process. Fuzzy Time Series (FTS) is a concept known as artificial intelligence which use to predict a problem where the actual data was formed in the values of linguistic. This study discusses the FTS method developed by Cheng to forecast the Composite Stock Price Index (CSPI) in October 2016. Within FTS, long intervals determined in beginning process. Based on FTS Cheng method with interval determination using frequency distribution, forecasting stock index based on data from January 2011-September 2016 result forecast for the month of October 2016 was 5.367.98 points. Based on calculation of MAPE, CSPI data from January 2011-September 2016 had an error value as big as 2.56% and has an accuracy of forecasting results amounted to 97.44%. Forecasting use the FTS Cheng has a great performance because it has MAPE value below 10%.