Correlation Query Processing Of Knowledge Graph Based On Bayesian Network

With the diversity of data sources and population of Web 2.0,data analysis,knowledge discovery/management are confronted with new challenges.Knowledge Graph(KG)provides a novel organization for massive data and domain knowledge.As the important foundation of information search and services in massive data background,correlation query processing is the subject with great attention and key technology in KG research.In this thesis,taking electronic commerce application as the background and Bayesian network(BN)as the underlying knowledge framework,we discuss correlation query processing of KG based on BN.We propose an effective method to fuse the domain knowledge in KG and the knowledge implied in users' behavior records to construct the BN that describes the correlation among commodities and the corresponding uncertainties.Then,we compute the indirect correlations of the commodities based on the BN inference mechanism.Our work can be summarized as follows:●With respect to the large-scale KG,we propose parallel algorithms for KG domain knowledge analysis based on Spark and GraphX to extract correlations among commodities.●With respect to massive users's behavior records,we propose parallel algorithms to extract correlations and fuse it with correlations in KG.we obtain a directed acyclic graph(D AG)for expressing correlations among commodities and compute the conditional probability table(CPT)for the DAG.●With respect to the DAG and CPT,we propose parralle algorithms to compute the indirect correlations of the commodities based on the BN inference mechanism and obtain the query results including correlative commodities and strength of correlations.Finally,with respect to the large-scale KG and massive users' behavior records,we give Spark-based parallel algorithms for BN construction and probabilistic inferences.Experimental results on real data show that the methods proposed in this paper are effective and efficient.