Monitoring and training stock prediction system for historical & live dataset using Lstm & Cnn

Abstract

A country development and stability are directly associated with its economy and today’s economy is profoundly dependent on the Stock market. Stock market indexes are subject to continuous change with respect to time, a hype or fall in the stock market has a crucial role in deciding the investor’s profit. Due to the economical ups & down and rapid growth in profit from the stock market, there required a need of developing a software application which continuously monitor the stock index’s and a prediction algorithm which can predict the possible change in stock index as for where it can go in future. Prediction of stock market does not follow any rules or predefined guidelines, hence prediction of stock market is difficult to achieve and the data-set for stock market prediction is also non-linear in nature which requires an efficient approach to resolve the time-series dependency of non-linear data. In our proposed system we are using the LSTM (long short-term memory) for efficiently predicting the stock index on historical data and the sudden change in stock market due to number of un-controllable factors is analysed by CNN model. As per the noise in the data-set we are employing wavelet denoising technique. If any changes in stock index with more than 10% of its initial value is analysed by monitoring module, then the system will notify the user with the change and also aggregating the result of predicting algorithm on that specific stock. Using our model Moneypred the accuracy in stock prediction is more than 70%.