NUS-ISS Learning Day 2018- Sentiment analysis in finance
- 1. Sentiment Analysis in Finance #ISSLearningDay2018 Eric Tham, NUS-ISS 13 July 2018
- 2. Content • Introduction • Sentiment is psychology • Demonstration in the Chinese equity markets • Conclusion #ISSLearningDay2018
- 3. A little about myself • Senior Lecturer in AI and analytics in NUS-ISS • 16 years of experience in the financial industry spanning risk management, quantitative analysis and data science. • 2 masters degree and completing Phd in Finance • Previous experience in building sentiment indices and trading strategies for the China equities in start-up in GZ • Teaches Blockchain, sentiment analysis and predictive analytics at NUS #ISSLearningDay2018
- 4. Previous work on Sentiment Analysis #ISSLearningDay2018
- 5. Introduction to Sentiment Analysis #ISSLearningDay2018
- 6. Alternative Data • New form of alternative data used in finance for trading use. • Traditionally, traders use technical analysis, news, prices, economic data to trade the markets. • Alternative data: • news and social media textual data • Satellite images, e-transactions, video etc. #ISSLearningDay2018
- 7. Sentiment Analysis • Two different sources <-> Different transmission impacts on financial markets • Main news media (Reuters, WSJ, CNBC, CNA ) • Peer to peer sharing • Twitter, Facebook, Weixin) • Specialised investment sites: SeekingAlpha and eToro, iMaibo in China • Big Data: wisdom of the crowd • Social media posts ‘peek’ into investor thoughts and beliefs #ISSLearningDay2018
- 8. Natural Language Processing • Systematically and computationally retrieves the public beliefs • Uses supervised learning – a reference bible that has positive or negative sentiment labelled • Use it to train a ‘A.I. model’ subsequently used to label other posts • The label peeks into what people are talking about the stocks, and also influences what others think. #ISSLearningDay2018
- 9. Simple Illustration of NLP • Systematically and computationally aggregates the public beliefs Some manual examples • https://seekingalpha.com/article/4186203-omega-healthcare-storm Positive or negative sentiment? • https://www.marketwatch.com/story/celgene-acceleron-shares-rise-on-blood-disorder-drug- study-2018-07-09?mod=BreakingNewsMain&link=sfmw_tw Positive or negative sentiment? #ISSLearningDay2018
- 10. Sentiment Indices • Aggregates across all posts and social media to obtain indices. Example (built by author) #ISSLearningDay2018
- 11. Sentiment is psychology Describes underlying concepts of sentiment in finance #ISSLearningDay2018
- 12. Behavioural finance • Basis of how you actually use sentiment correctly? • The market is about psychology. Therein, why sentiment works. • Over-reaction to attention-seeking news. • Under-reaction to important news due to transmission mechanism • Investor disagreement in market. • Do back-testing on results for best trading strategy in the markets. #ISSLearningDay2018
- 13. Sentiment trading Strategies • Mean reversion : • out of over-reaction? • Or under-reaction? • Momentum • Irrational exuberance? • How do you detect momentum from sentiment? • Different in different markets #ISSLearningDay2018
- 14. Demonstration for China equities Section description #ISSLearningDay2018
- 15. Trading Strategies: Sentiment HeatMap #ISSLearningDay2018 What do the colors and sizes indicate?
- 16. Some references • Trading on Sentiment • Handbook of News Analytics in Finance #ISSLearningDay2018
- 17. Some references • Author’s previous work • http://forport.tzmaibo.com/docs/imaibowhite_2.pdf • http://forport.tzmaibo.com/docs/imaibowhite_1.pdf #ISSLearningDay2018
- 18. Conclusion • Markets about psychology. • Sentiment from textual data a Big Data way to perceive this market psychology. • This sentiment can be used for trading strategies but needs systematic back-testing and a good understanding of the markets still #ISSLearningDay2018