Prediction of Investor-Specific Trading Trends in South Korean Stock Markets Using a BiLSTM Prediction Model Based on Sentiment Analysis of Financial News Articles

Abstract

Stock market performance is determined by supply and demand of individual, institutional, and foreign investors, who increasingly use media such as news articles for decision-making. We present a bidirectional long short term memory model to forecast trading trends based on statistically significant investor-specific topics from financial news datasets. The application of this study shows three valuable results: (i) topics significantly meaningful to each investor type differ, (ii) investors show different decision-making trends for the same news topics and different sensitivity levels, and (iii) news topics significantly associated with investors’ responses differ according to the stock market and sensitivity.

Keywords:

• Sentiment analysis
• Bidirectional long short-term memory
• Investor behavior
• Artificial neural networks
• Topic modeling

Data availability

The data that support the findings of this study are openly available in Naver Financial at https://finance.naver.com/news/mainnews.nhn (financial news) and https://finance.naver.com/sise/sise_trans_style.nhn (investors’ trading trends).

Disclosure statement

No potential conflict of interest was reported by the authors.