Reinforcement learning based predictive analytics framework for survival in stock market

摘要

Contemporary research in stock market domain is limited to forecasting of the stock price from one day to one week. Such small period predictions cannot be of much help for continuous gainful survival in the stock market. In fact, there has to be predictive analytics framework which analyses the current situation in the holistic manner and provides the appropriate advice for selling/buying/no action along with the quantity resulting in significant gain for the user/investor. The proposed framework generates various reinforcement signals by applying statistical and machine learning techniques on historical data and studies their impact on the stock prices by analysing future data. The outcome of the process has been used to generate rewards, through the use of fuzzy logic, for various actions in a given state of the environment. Fully automated implementation of the proposed framework can help both institutional and common investor in taking the rational decision.