Explainable Machine Learning Exploiting News and Domain-Specific Lexicon for Stock Market Forecasting

摘要: In this manuscript, we propose a Machine Learning approach to tackle binary classification problem whose goal is predict the magnitude (high or low) of future stock price variations for individual companies S&P 500 index. Sets lexicons are generated from globally published articles with identifying most impactful words on market in specific time interval and within certain business sector. A feature engineering process then performed out lexicons, obtained features fed Decision Tree classifier. The predicted label represents underlying company’s variation next day, being either higher lower than threshold. performance evaluation have carried through walk-forward strategy, against set solid baselines, shows that our clearly outperforms competitors. Moreover, devised Artificial Intelligence (AI) explainable, sense analyze white-box behind classifier provide explanations results.