Alternative data has rapidly gained popularity as a means of observing investor reactions to current stock market events, analyzing sector movements, and developing investment strategies. Signals observed from alternative data may add valuable information to conventional market analysis, especially when there exists some information asymmetry in the market, to which small-cap stocks are particularly sensitive. Our research uses alternative textual data obtained from social media to predict the price momentum of small-cap stocks, which are known to be sensitive to market events and retail investors’ reactions. We directly address the challenges in the analysis of texts, especially in the Korean language, such as the use of jargon, acronyms, and typos, by taking a two-fold approach to preprocessing. Then, we extract the social keywords and label them by two standards: word momentum and correlation with stock price movement. Using these social keywords as social indicators derived from the retail investors trading small-cap stocks, in addition to the conventional technical indicators, we propose a price momentum forecast model which classifies whether the closing price of a given small-cap stock will increase by more than 10% in the next 5 business days. Experimental results show that our approach successfully extracts social keywords that play a significant role in determining the price momentum in the market, and our selected social indicators are intuitive to human understanding. Furthermore, we consider example investment scenarios and run tests by training various classifiers with the social indicators and computing cumulative returns. Our results are very promising and show that the investment scenarios with social indicators outperform the benchmark considerably. We anticipate that our work will expand the scope of using alternative data for stock market analysis and price momentum forecasts by incorporating social indicators extracted from social media.
Read full abstract