Nonlinear asset pricing in Chinese stock market: A deep learning approach

The redesign of asset pricing models failed to integrate the frequent financial phenomenon that stock markets exhibit a non-linear long- and short-term memory structure. difficulty lies in developing nonlinear structure capable depicting influence variable. This paper presents Long- Short-Term Memory Neural Network Model (LSTM) capture among five elements Chinese market, including market portfolio return, capitalisation, book-to-market ratio, earnings factor, investment factor. long–short-term implies autocorrelation function return series decays slowly has long-term characteristic. LSTM model surpasses standard Fama–French five-factor terms out-of-sample goodness-of-fit long–short strategy performance. empirical findings indicate properly represents relationships between components.