Study On Intelligent Matching Techniques Of Transportation Ontologies

The data of different intelligent transportation systems have the characteristics of multi-source and strong heterogeneity,which hinders the interaction and sharing between them.As the core of semantic technology,transportation ontologies provide a method for knowledge representation and describe domain knowledge by formally defining relevant concepts in the transportation domain and the relationship between these concepts,and provide effective technical means for data and knowledge representation,management and sharing for the cooperation between intelligent transportation systems.However,different transportation ontologies also have the problem of data heterogeneity.In order to promote the interaction and cooperation between intelligent transportation system applications,it is necessary to carry out the transportation ontology matching process to determine the semantic relationship between heterogeneous concepts and realize the interoperability of different intelligent transportation systems.Therefore,this paper mainly studies the intelligent matching techniques of transportation ontology,focusing on three key issues: how to automatically optimize the weights of different similarity measures to improve the quality of transportation ontology alignments,how to adaptively build an effective similarity measure model to accurately identify heterogeneous concepts,and how to efficiently determine the entity set in large-scale transportation ontology:(1)Aiming at the first key issue,the single objective optimization model of transportation ontology meta-matching is established,and three approximate assessment metrics are proposed.On this basis,a UCGA(Uniform Compact Genetic Algorithm)based matching technique is proposed,and the detailed steps of solving the single objective optimization model with UCGA algorithm are given: coding mechanism,probability vector,linearly reduced virtual population,and local search strategy.The experimental results show that UCGA can effectively match various transportation ontologies with diverse heterogeneities.(2)Aiming at the first key issue,a multi-objective optimization model of transportation ontology meta-matching is established,whose objectives are the proposed approximate recall and precision.An I-IM-MOEA(Improved MultiObjective Evolutionary Algorithm with Inverse Model)based matching technique is proposed and the detailed steps of solving the multi-objective optimization model with I-IM-MOEA are described: coding mechanism,adjusted selection operator,dynamic reference vectors,and local search strategy.The experimental results show that the proposed technique based on I-IM-MOEA can more effectively match all kinds of transportation ontologies.(3)For the second key issue,a similarity measure based on multi-level context information is proposed to effectively identify the heterogeneous entity correspondences of transportation ontology.Then,for the third key issue,a single objective optimization model of transportation ontology entity matching is established,and a pc DE-ASL(parallel compact Differential Evolution with Adaptive Step Length)based matching technique is proposed to improve the efficiency and quality of alignments.The experimental results show that the proposed similarity measure based on multi-level context information can better identify various heterogeneous concepts than the cutting-edge similarity measures,and the quality of ontology matching results obtained based on pc DE-ASL ontology matching method can match various transportation ontologies more effectively and efficiently than the cutting-edge ontology matching techniques.