Research On Domain Ontology-based CBR Optimization And Its Application

In the Artificial Intelligence field, due to acquiring knowledge easily and robust reasoning, Case-Based Reasoning (CBR) compensates traditional expert systems’ defects such as unable to deal with mass data. Moreover, CBR has been applied practically and successfully to problem solving and decision support domain. However, CBR faces with following challenges which impact its current development pace:knowledge cannot be shared and integrated caused by different case representation model and knowledge storage in cross system which leads to Information Silos. Meanwhile, as lack of expansibility in case representation and reusability in case reasoning strategy, it’s difficult to update and evolve CBR system resulting in reason results not fully meet user needs.Based on the above issues, we study academic achievements and application status in CBR field and we find ontology which is able to provide explicit and shareable specification of a conceptualization, can make up the defects of CBR fundamentally. Hence, it is inevitability to combine ontology with CBR. After studying and analyzing on ontology, we also find that in ontology classification, domain ontology not only owns ontology’s conventional advantage in knowledge representation, but also is rich in detail and specific domain knowledge which can be used to improve reasoning more effectively. Thus we emphasize on researching on domain ontology-based CBR.Based on CBR process model, we explain the principles and methods of how to combine domain ontology with CBR. On the one hand, using of knowledge model from domain ontology helps CBR to improve case representation such as lack of standardization and expansibility. On the other hand, extracting domain knowledge from domain ontology optimizes CBR similarity calculation, as well as assists case revise and retain in satisfiability test. Overall, domain ontology can promote the entire process of CBR. Finally,2e build domain ontology-based CBR system architecture in reference to CBR task model for practical applying.In order to verify the effectiveness and feasibility of these proposals, we use protege and myCBR tool to develop recipe recommendation system in rapid prototyping. The main task includes:building recipe ontology based on Chinese Cuisine Culture ontology, recipe representation and similarity algorithm strategy design of recipe recommendation system and so on. By comparison with other recipes retrieval system, it approves that domain ontology based recipe recommendation CBR system performed more effective in recommending appropriate recipe results.