Stock Price Prediction using Reinforcement Learning and Feature Extraction

Abstract

The goal of this project is to develop a new technique to predict stock worth’s through the usage of Reinforcement Learning & Sentiment analysis from social media. During this paper we are going to analyze economical technique which may predict stock movement accurately using both Historical & Real-time Data. The Q Learning based approach will be used to predict these Stocks over a Partially Observable Markov Decision Process comprising of any number of Stocks taken as a State & providing 3 actions which are Hold, Sell or Buy. Additionally, Social media offers a robust outlet for people’s thoughts and feelings it's a fast-ever-growing supply of texts starting from everyday observations to concerned discussions. Using social media comments & tweets analysis regarding a Stock, an efficient data can be obtained which can help in determining the overall public review of the Stock. Using these two distinctive approaches together, an efficient technique can be developed to predict Stocks Prices with more accuracy.