Bayesian Network Structure Learning Under High Dimensional Data

As a traditional and effective probability graph model,Bayesian network has been widely studied by scholars because of its causal and probabilistic semantics.Under high-dimensional data,learning the Bayesian network structure with traditional algorithms not only consumes a lot of time,but also the complexity of the network grows exponentially with the increase of the data dimensionThis paper proposes a Bayesian network structure learning algorithm for high-dimensional data,LTB algorithm.The algorithm is combined with Lasso(Least absolute shrinkage and selection operator),Tabu Search algorithm and BIC(Bayesian Information Criterion),which mainly solves the problem that the Bayesian network structure is diffi-cult to analyze.The main steps of the algorithm are as follows:1.Use Lasso to reduce the dimension of the independent variable,and filter out the independent variable closely related to the response variable as the apex of the Bayesian network;2.Select the tabu search algorithm as the meta heuristic algorithm,and BIC is chosen as the method of calculating the score,and the two are combined to construct the global optimal Bayesian network structure.The example shows that the tabu search algorithm has significant ad-vantages in learning the accuracy of Bayesian network structure,and the proposed LTB algorithm has significant advantages in running timeThis paper takes the influencing factors of Shanghai Securities Composite Index as the analysis object,applies the proposed LTB algorithm to learn the data of the factors affecting the Shanghai Composite Index,and obtains the Bayesian network structure that affects the monthly yield of the Shanghai Composite Index.The causal relationship be-tween Shanghai Securities Composite Index and its influencing factors can also use the conditional probability to obtain an effective way to regulate the monthly yield of the Shanghai Composite Index.Finally,through analysis,it is concluded that correctly guid-ing irrational traders to trade stocks,increasing demand deposits of securities companies or fund management companies,increasing consumer confidence index,and increasing stock turnover are all effective ways to increase monthly stock returns.