Bayesian network structure optimization method and device based on frequent item mining

Abstract

The invention relates to a Bayesian network structure optimization method based on frequent item mining. The method comprises the following steps: acquiring a data set; constructing an association rule set according to association rules reflecting the association degree between the random variables; after frequent item sets are extracted from the data set, learning a preset Bayesian network, and obtaining a maximum Bayesian network structure set corresponding to the maximum frequent item set; according to the association rule set and the maximum Bayesian network structure set, extracting a white list and a black list from the data set, and then constructing penalty terms; and according to the penalty terms and the BDeu scoring function, obtaining a scoring function fused with prior, and then carrying out loop search by using a hill-climbing search algorithm to obtain an optimal Bayesian network structure. By adopting the method, the efficiency and accuracy of Bayesian network structure learning can be obviously improved.

Description

technical field [0001] The present application relates to the field of data processing, in particular to a Bayesian network structure optimization method and device based on frequent item mining. Background technique [0002] With the development of technology in the field of machine learning, Bayesian network has emerged. Bayesian network is a general probability graphical model, which can represent the correlation between random variables and is a graph of the joint probability distribution of random variables. Therefore, it has strong interpretability. Bayesian network has a wide range of applications in the field of machine learning because it can represent uncertain knowledge and perform inference calculations. It is one of the hot spots in the field of research. [0003] However, in real machine learning problems, due to the noise and size limitations of actual sample data and the complexity of network space search, learning the structure of Bayesian networks is a non-...

AI Technical Summary

Problems solved by technology

These studies focus on the interpretability of the model to the training data. The learning accuracy is mainly determined by the quality of the training data. In the case of limited data or complex problems, overfitting often occurs and it is difficult to apply in practice.

Images