Paper Number
1960
Paper Type
Completed
Description
Investor psychology provides an important avenue for modeling non-fundamental behaviors in financial analysis. Yet, whether market psychological information has a practical application in predicting asset returns is still under debate. Thus, a burgeoning number of machine learning algorithms have been developed to test the effectiveness of investor psychology in financial predictions. With all the merits of machine learning approach, the drawbacks are prediction biases, data overfitting issues and poor performance. To address these issues, we developed a DeepPsych system to harness the power of high frequency TRMI psychology data for market prediction. In a “hybridization–generalization–dual-channel-fusion” three-stage experiment, we evaluate each proposed module and the entire framework against the state-of-art machine learning benchmarks on investor psychology and trading data of the SPY (SP500 ETF). Results demonstrate that our deep learning framework can automatically identify features that are more effective than fundamental factors and support profitable trading.
Recommended Citation
Shen, Jiancheng; Wang, Jia; Zhu, Hongwei; and Liu, Benyuan, "DeepPsych: Harnessing Market Psychology with Deep Learning" (2023). ICIS 2023 Proceedings. 17.
https://aisel.aisnet.org/icis2023/dab_sc/dab_sc/17
DeepPsych: Harnessing Market Psychology with Deep Learning
Investor psychology provides an important avenue for modeling non-fundamental behaviors in financial analysis. Yet, whether market psychological information has a practical application in predicting asset returns is still under debate. Thus, a burgeoning number of machine learning algorithms have been developed to test the effectiveness of investor psychology in financial predictions. With all the merits of machine learning approach, the drawbacks are prediction biases, data overfitting issues and poor performance. To address these issues, we developed a DeepPsych system to harness the power of high frequency TRMI psychology data for market prediction. In a “hybridization–generalization–dual-channel-fusion” three-stage experiment, we evaluate each proposed module and the entire framework against the state-of-art machine learning benchmarks on investor psychology and trading data of the SPY (SP500 ETF). Results demonstrate that our deep learning framework can automatically identify features that are more effective than fundamental factors and support profitable trading.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
13-DataAnalytics