CAPM Based On Stock Correlation Network

The CAPM is proposed on the basis of portfolio theory, but the use of the model is based on some assumptions, and we do not consider the correlation between the stocks when we calculate the expected return of a stock or a portfolio with the CAPM. Eventually it leads to that its applicability in today’s stock market, especially in the Chinese market is not very high. Because stock correlation network is established based on the correlation among the stocks, we use the stock correlation network to optimize the CAPM in this thesis.This article selects 285 closing price data of 2446 stocks from December 2,2013 to December 31,2014 for the research. Due to the huge of the stock market, we first use the optimal clustering number to divide the stock market into several classes, and then establish stock correlation network based on the stock return correlation coefficient for each class. We conduct centrality analysis for stock correlation network to optimize the CAPM. Finally, we use the simulated data for validation.The main conclusion is that it not only encourages people to focus on the interesting stock correlation network but also makes the optimized CAPM forecast MSE smaller by using the optimal clustering number and the stock correlation network, that is to say, the forecast effect is better. Thus, we think that the optimized CAPM can weaken the impact of related factor which influences the classical CAPM prediction effect to a certain extent.