Financial markets are event-driven trade places, and news is one of the modes of communicating information about these events. A trader forms an expectation about market risks and returns from the market information, and hence takes a trade position. This study captures traders’ trade positions and trade directions at the bid-ask stage of the futures market, and studies the impact of high-frequency news on traders’ trade direction in the futures market.
Sentiment analysis for the financial markets focusses on quantifying the nature of the impact of the textual market information on the market sentiment. This study is one of the first sentiment analysis research works conducted on futures markets and carries implications on how news impacts the futures market trade directions. The methodology adopted in this work is hybrid sentiment analysis – wherein both, the dictionary-based sentiment and the machine learning-based sentiment analysis, are used to predict the market trade directions or sentiment. The study uses Loughran and McDonald’s financial sentiment dictionary for the dictionary-based sentiment analysis, and news headlines-based vector space model (VSM) trained using support vector machines (SVM) for the supervised sentiment analysis. The study conducts a high-frequency news-based sentiment analysis and has implications on algorithmic trading positions on the futures markets.
The machine learning-based sentiment analysis entails the process of news-trend alignment to create automatically labelled training data – which is one of the most critical steps in building the accuracy of the model. Previous works on machine learning-based sentiment analysis empirically established an optimal alignment lag for a robust training data in the stock market context. This study establishes that the arbitrage window of opportunity in the Indian futures markets is as short as five minutes, which underlines the importance of high-frequency analysis of news on the markets.
