An Empirical Study of Stock Return and Investor Sentiment Based on Text Mining and LSTM

DOI

10.2991/icssed-19.2019.104How to use a DOI?

Keywords

Investor sentiment; natural language processing; text sentiment analysis; deep learning; LSTM; stock return; expanded asset pricing model.

Abstract

Based on the development of social network and big data, we adopt the unstructured text-based investors’ comment data mining from the stock bar forum, and use long short-term memory neural network for text sentiment analysis to build a more accurate investor sentiment indicator. Based on this indicator, an empirical study on the component stocks of the GEM Composite Index is conducted to explore the impact of investor sentiment on stock return. Through a full sample stock selection test, we find that the performance of the portfolio based on investor sentiment indicator performs significantly better than the benchmark. Further more, compared with the basic Fama-French three factor model, the goodness of fit and significance of the asset pricing model with investor sentiment factor added are both improved, indicating that the investor sentiment index we constructed can capture the investors’ sentiment in the market well, and has a good explanatory power for stock return.

Copyright

© 2019, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Ren Tianyu
PY  - 2019/05
DA  - 2019/05
TI  - An Empirical Study of Stock Return and Investor Sentiment Based on Text Mining and LSTM
BT  - Proceedings of the 2019 4th International Conference on Social Sciences and Economic Development (ICSSED 2019)
PB  - Atlantis Press
SP  - 554
EP  - 558
SN  - 2352-5398
UR  - https://doi.org/10.2991/icssed-19.2019.104
DO  - 10.2991/icssed-19.2019.104
ID  - Tianyu2019/05
ER  -