Bayesian network inference method based on beetle antennae strategy

Abstract

The invention discloses a Bayesian network inference method based on a beetle antennae strategy, and aims to solve the problems that some of the most probable explanation methods of Bayesian network inference are low in search efficiency and are easy to fall into local optimum. A framework of a population strategy is applied to Bayesian network inference, a most probable explanation method framework is established, a longicorn population is used for optimizing possible explanation, a generation method for establishing a dominant initial population according to a chaos strategy is provided, population diversity in an initial stage is enhanced, a genetic operator is introduced to construct a direction combination sequence of search behaviors at the same time, according to appropriate level pheromones, self-adaptive dynamic parameters are established to adjust the moving step length of detection behaviors so as to improve beetle antennae detection behaviors, and through population iteration, the optimal most possible explanation of the Bayesian network is finally obtained, so that the inference rate and results of the Bayesian network are improved.

Description

technical field [0001] The invention relates to the field of data mining based on probability theory and graph theory, in particular to a Bayesian network reasoning method based on the longhorn strategy. Background technique [0002] Currently, Bayesian network reasoning is most likely to be explained in many fields, such as industrial fault diagnosis, sociological statistics, weather forecasting, economic model calculation, etc. Bayesian networks are the main tool for probabilistic reasoning, which give a complete joint probability distribution and use directed edges to represent the dependencies between variables. Bayesian network reasoning is the ultimate goal of applying Bayesian networks. Computing the most likely explanation of Bayesian networks is one of the main problems solved by Bayesian network reasoning. The most likely explanation is a set of given Bayesian networks. In the case of observed variables (evidence nodes), the state of all non-observed variables (no...

AI Technical Summary

Problems solved by technology

[0003] Solving the most probable explanation of the Bayesian network is proved to be a non-deterministic problem with polynomial complexity. The current methods for solving this problem are mainly divided into two types: query-based methods and local search-based methods. The query efficiency is higher in a small network, but it is difficult to find an effective solution in a network with a large number of nodes.

However, methods based on local search can find approximate solutions or optimal solutions in networks with a large number of nodes, but many of these methods are prone to fall into local values, and the search efficiency needs to be improved.

Images