Research On Theft Cases Inference Based On Markov Logic Networks

With the rapid development of economy and information technology in recent years,the criminal theft cases have continued to occur at a high rate.Theft cases not only affect the quality of life of urban residents,but also interfere with public order.Due to the extremely limited police resources of the public security departments and the high frequency of theft cases,the theft cases are in a state of high incidence and low breaking.How to use data mining technology to improve the detection rate of theft and the quality of life of urban residents under limited police resources is a problem facing the public security department.This paper proposes a theft cases inference model with association rules analysis and Markov Logic Networks.Markov Logic Networks can realize reasoning by incorporating rule into domain knowledge,but it needs the participation of a lot of domain experts to give rule set,and domain expert can not guarantee all effective rules.The entry point for this paper is to learn domain rules from domain data by using relevant algorithms without the involvement of a large number of domain experts.Firstly,two basic assumptions are put forward: one is that all the theft cases can be expressed by domain concept vector;the other one is that all the rules are composed of the causality of domain concept.Based on these two assumptions,this paper first uses the ontology thought,constructs the core concept classification system of the theft cases,takes this system as the standard,expresses all the theft cases as the concept vector set;Secondly,using association rules mining algorithms for inference to find the strong rules in the set of concept vectors.This paper adopts the Apriori algorithm based on effect to realize the quadratic effective filtering of strong rules,obtaining the strong rules of the concept causality composition.In this paper,an approximate computing method bases on adjoining matrix of administrative divisions is proposed under the circumstance of high computational complexity of calculating case distance.Finally,this paper considers the characteristic that Markov Logic Networks can include domain knowledge succinctly,using this model to realize domain knowledge reasoning,the rules of the theft cases are used as the input interface of the rules of Markov Logic Networks,constructing Markov Logic Networks and finally realize the reasoning of theft cases.The experimental results show that the Markov Logic Networks model based on association rule learning has a significant improvement in accuracy and performance compared with the empirical rule based Markov Logic Networks model.When top-K is set to top-30,the highest accuracy can be increased by 11.4%.In terms of performance,the optimization effect of reasoning time increases linearly with the increase of case data.