Stock value prediction using machine learning

Abstract

The prediction of stock value has always been esoteric to stock analysts and statisticians. However, the stock market banks on public investment for its survival which also accounts for its dynamic nature. Previous methods of stock value prediction involve implementation of applied statistics, machine learning, news feed extraction with a moderate prediction accuracy. The proposed system involves the best practices from previous attempts and also a new approach to stock value prediction which would have an improved prediction accuracy than previous systems. The proposed system is implemented using deep learning: LSTM (Long short-term memory) and RNN (Recurrent neural network) algorithms which act as the prediction model and thus helps in delivering accurate predictions for the future by analyzing the pattern of variable stock prices for a time period. A conjunctive system of a keyword extractor and a sentiment analyzer directed towards news articles hosted by Twitter would help indicate the current performance of the company whether optimistic or not. The usage of deep learning algorithms provides a more robust mechanism to predict stock prices. The sentiment analyzer indicting the performance of the company thus acting as an important asset for investors to understand the stability of the company during the long term. The proposed system holistically covers all the important parameters considerable for an investor to invest in a particular company. Also, the proposed system helps in eliminating the esoteric nature behind stock analysis and encourages the common investors with partial knowledge of finance to invest in the stock market.