Research Projects

The study aims to analyze the evidence of information bias, leverage effects, and spill over of conditional volatility in international stock market returns using machine learning techniques like Long-Short-Term-Memory network and deep learning. It proposes using sign bias test and asymmetric GARCH models to capture the existence of leverage effects on conditional volatility among international stock market indices. The multivariate generalized autoregressive conditional heteroskedasticity (MGARCH) model is also proposed to address volatility spillover effects. The study also proposes using a non-linear asset pricing model with deep neural networks for a group of international stock markets clustered based on geographical boundaries, market turnover, and liquidity. The originality of the study is to use advanced univariate and multivariate time series modes and stochastic no-arbitrage condition of neural network algorithm to decode the market return and its volatility pattern. The stock market indices will be collected from subscribed sources of Bloomberg. The study will use high frequency hourly and daily data of the international stock markets. The inclusion of no-arbitrage and stochastic constraint in the learning algorithm may significantly capture the market dynamics and pattern of return volatility among international stock markets.