Forecasting Daily Close Prices of Stock Indices using LSTM

The stock market remains one of the most un-predictable and volatile activity, numerous arithmetical and computational methods have been designed to get the maximum reap from it. A lot of predictive methods and models have been designed in the recent years such as Linear Regression, Multiple Regression and RNN’s to help traders, analysts and investors make the most probabilistic decision. Here we use RNN’s Long Short-Term Memory architecture, a purpose oriented RNN to predict the bear market tables. Statistical techniques such as linear regression, polynomial regression are popular techniques for traders to see the association between variables and hence predict stocks. Using RNN’s and LSTM gives a better understanding of the changing prices seen during trading of equity to day traders and financial advisers since LSTM is a beautiful and efficient way of dealing with time stepped data.A lot of money has been made and lost by capitalizing the stock market, prediction models are not accurate but specify a more calculated approach to make money. By no means do any of the financial models or predictive models guarantee the making of money.This project aims to forecast the prices of the Apple stock from the year 2014-2019 and analyze the day to day change in the value, using LSTM in keras and the backend tensor flow, from the data extracted from yahoo.com. The data is then split into training and testing datasets, where the test data is our predicted value of the stock. This model, once accomplished shows us that for day to day analysis, this can be an appropriate alternative to other models of forecasting.Keeping in mind the potential of the project, the future scope including applications using the project has been discussed in this paper.