基于DCC-MGARCH模型的股票网络构建

Stock Network Building Based on the DCC-MGARCH Model

针对股票的相关性,首先建立指数平滑预测模型,对股票的周开盘价和周收盘价进行预测,把缺少的个股回报率数据补齐,运用DCC-MGARCH模型计算出多只证券之间的动态相关系数,并运用granger检验法对多只证券的收益率与资产折现率进行因果关系检验。根据总相关系数矩阵,设定不同阈值,结合中心节点度最大化和股票网络平均路径最小化两个原则选取最优的股票网络,以期为投资者提供投资以及股票市场板块划分参考。

According to the correlation of stocks, exponential smoothing forecasting model is established to forecast the opening and closing week prices of stocks. With the supplement of some stock return data, the DCC-MGARCH model is used to calculate the dynamic correlation coefficient among a variety of securities, and the Granger inspection method is applied for conducting causality test on the securities＇ return rate and assets discount rate. According to the general correlation coefficient matrix, different thresholds are set; combined with the two principles of maximum central node and minimum average path of stock network, the optimal stock network is selected, in order to provide valuable reference for investors＇ decision and the plate division of stock market.