Effect of exogenous market sentiment indicators in stock price direction prediction

As the most popular methodology in stock price prediction, technical analysis accounts for 66 % of stock price prediction studies for 11 years’ period (2007 ∼ 2018), meanwhile fundamental analysis and combined analysis using technical and fundamental analysis are just 23 % and 11 % of studies, respectively. Technical stock price analysis generally expands the feature space by generating price-derived technical indicators from daily stock prices. However, using price-derived technical indicators is criticized for its complete subjectivity and self-fulfilling prophecy. As a remedial action for price-derived technical indicators, this study includes three popular exogenous market sentiment indicators (EMIs): market volatility index (VIX), fear-greed index (FGX), and put-call ratio (PCR), and investigates the importance and effect of the selected EMIs in stock price direction prediction. To capture the maximal effect of the EMIs in stock price direction prediction, the best feature subset (BFS) selection method is adopted in company with the extreme gradient boosting (XGB) algorithm, which significantly improves the prediction accuracy and interpretability with fewer input features. The proposed BFS with XGB method adequately reflects the collective behaviour of the EMIs with price-derived technical indicators of time series stock price data. Additionally, combined with a piecewise time segmentation technique , the piecewise best feature subset (PBFS) selection method shows that the three EMIs: VIX, FGX, and PCR make a considerable effect in prediction accuracy and interpretation of stock price direction prediction.