An Automated Framework for Incorporating Fine- Grained News Data into S&P BSE SENSEX Stock Trading Strategies

Objective: In this paper, we purpose a framework for automatic incorporation of fine-grained news data to generate S&P BSE Sensex trading strategies. Method/Statistical Analysis: Financial news containing information about companies events, share prices, sales, Government and economic policies etc., are fetched from web. For fine-grained news data, text mining is applied to pre-process the news. Sentiments analysis is done to estimate the exact sentiment of news messages containing contradictory words and input at right place into the framework. News entities specific to stock market behavior prediction are extracted from fine-grained news data. Pearson Correlation characterizes the relation between stock returns and expert defined Impact Factor (IF). Findings: The purposed framework incorporates the news dataset specific to 30 companies listed under S&P BSE Sensex stock index. Stock returns corresponding to news entities generates the buy or sell signal. The results depicting positive value of Pearson’s correlation describes that market moving in right direction. It is proved that the specified news entities incorporated at right place into the framework have significant effect on stock movement and generates higher magnitude of returns as compared to other entities used in previous work. The framework gives approximately 94% accuracy which is greater than other news based trading model. Application/Improvements: The accuracy of purposed framework can be improved by combining a number of technical trading indicators with news entities. Also optimal trading strategies can be generated by using optimal programming.