Indices prediction of Bangladeshi stock by using time series forecasting and performance analysis

Prediction of share index is a very mystical task for the traders who take part on stock business. Investors invest their money to earn profit from this sector. However often they face misfortune in their life due to the wrong prognosis of stock index. In this work, we have predicted the indices of Bangladeshi stock by using various stock prediction algorithms such as Feed-Forward Neural Network (FFNN), Auto Regressive Integrated Moving Average (ARIMA), Linear model & Holt- Winter approaches and analyzed the performance of these algorithms over 35 Bangladeshi stocks. Time series analysis is considered to this work and performance of algorithms is computed by calculating percentage accurate prediction. From analysis, it is found that, ARIMA (1,0,0) gives maximum prediction accuracy (82.1%) in average among all and FFNN shows best algorithm in forecasting stocks index. FFNN gives maximum accuracy in 14 out of 35 stocks.