基于马尔可夫毯时序回归模型的房地产板块指数预测

房地产板块是中国股市的核心行业板块，房地产板块股价指数走势的分析有助于对中国股市态势的正确把握. 针对行业板块之间的相关性难以发现和房地产板块股价指数难以有效预测的问题，基于股市行业板块之间指数波动的相关性，利用马尔可夫毯学习算法选择与房地产板块相关的行业板块，通过因果分析避免现行方法中选取相关板块的主观性；进而，利用Granger因果检验从中选择与房地产板块存在时序因果关系的板块，从而构建房地产板块的向量时序回归模型，实现对房地产板块股价指数的有效预测；最后，通过脉冲响应和方差分解对模型进行分析. 对于上证股市所进行的实验比较和实证分析的结果表明，该方法能有效预测房地产板块股价指数的走势.

Real estate plate stock index predict based on Markov blanket timing sequence regression

The real estate plate plays an important role in China's stock market, the real estate plate stock price index trend analysis helps to understand the Chinese stock market situation correctly. For the correlation between industry plates is difficult to find in the stock market and the real estate plate stock price index is hard to predict. Based on the stock market index fluctuation correlations between plates, the industry plates correlated with the real estate plate were selected by using Markov blanket algorithm. Causal analysis method was used to avoid the subjectivity of selecting relevant plate in current methods. Then from those plates above we found which plate had sequential causality with the real estate plate by Granger causality test, established timing sequence regression model with real estate plate to predict real estate plate stock price index effectively. In the end the model was analyzed by impulse response and variance decomposition. For the Shanghai stock market the experimental comparison and empirical analysis of the results showed that, this method can effectively predict the real estate plate stock index trend.