Kalpit Munot

Forecasting stock markets with great accuracy is key to determining profits in the market. Training a neural network with previous values and then simulating this network could give predictions, but many previous attempts have had very low accuracy. Therefore de-noising the data and applying a HP filter (Hodrick-Prescott filter) will increase the accuracy of ANN (artificial neural network). The neural network used here is a recurrent neural network which is usually the best neural network for time series. Factors which are being considered here are opening, closing, maximum, and minimum prices. Data which is used here is obtained from yahoo finance which will be normalized and then be used for training our neural network.We will also compare our predictions with traditional neural networks (Simple back propagation network or feed forward network).