The Connectedness of Business Sectors and Return Predictions by a Hybrid Machine Learning Method

According to the Granger-causality test, this study investigates the causation of business sector return in the Thailand stock market. Moreover, the result reveals that sectors' return connectedness contributes to the predictive model for forecasting price. First, we construct vector autoregression (VAR) to extract the intersection causation. Next, the hybrid machine learning method, CNN-LSTM, is proposed and modified to make a prediction based on the linkage assumption. The finding shows that a given sector influences other sectors, and the CNN-LSTM, which incorporates stated assumption, gains better performance than plain LSTM. Consequently, we found supporting evidence for the assumption that there is a linkage of return where money moves around from asset to asset, as stated in the network theory in finance. Therefore, the forecasting model gains a better performance by adding this finding