Machine learning for asset pricing in digital finance

Asset pricing is an essential financial activity, which aims to predict the prices or values of claims to uncertain payments. Machine learning is a promising technology to predict asset prices as it has achieved impressive performance in some key domains (e.g., computer vision and natural language processing). The advantage of machine learning for asset pricing is that it can capture the non-linear attribute, considerable predictor variables, and rich specifications of the functional form. This has motivated several recent studies that use machine learning for asset pricing prediction. However, existing works ignore the noises in asset pricing data and assume that factors in asset pricing are equally important.

To improve the prediction accuracy of asset pricing, this project develops an attentional deep neural network. This new algorithm not only considers data noises, but also discriminates the importance of factors when predicting asset pricing.

This is an ongoing research project started in 2019, and is aligned with both the Digital Marketing and Finance and Artificial Intelligence, Machine Learning and Data Analytics research themes of the IDT.

For further info, please contact Dr Xiang Li.