Profiting from High Frequency Market Psychology Data with Deep Learning

It is challenging to predict financial markets, but there have been continued efforts to develop improved prediction methods. With availability of high frequency market psychology data, and guided by design science principles, this research iteratively develops and comprehensively evaluates DeepPsych, a deep learning system that leverages market psychology data to gain prediction advantage. Using two convolutional sequence-to-sequence channels to extract local and temporal features from psychology and trading data separately, the system outperforms other leading machine learning and deep learning models in both machine learning metrics and economic values realized through trading strategy based on the prediction. This research contributes to both information systems design science through innovation in deep learning and finance by providing empirical evidence about the predictive power of high frequency market psychology data. The research also benefits practice by producing a validated Fintech artifact.