Performance Analysis of Different Recurrent Neural Network Architectures and Classical Statistical Model for Financial Forecasting: A Case Study on Dhaka Stock Exchange

In recent years the advancement in neural network architecture and introduction of recurrent neural network has attracted a lot of interest to work with sequence data. LSTM is derived from the basic architecture of Recurrent Neural Network. It has memory units which extends the power of Recurrent Neural Network. In this paper, we analyze the performance of different advanced neural network architectures and classical time series forecasting method, e.g., ARIMA on selective stock prices from Dhaka Stock Exchange (DSE). Our experimental results show that the neural network models perform better than the ARIMA model in reducing RMSE (Root Mean Square Error).