The Research Of Case-based Reasoning Based On Intelligent Optimization Algorithms

The building of case base and case retrieval are the key issues in case-based reasoning and they influence the solving of the cases. The existing researches in case-based reasoning focus mainly on case retrieval. But a well built case base can not only improve the accuracy but also improve the efficiency in the process of reasoning. And in the exiting researches on case retrieval, the most popular distance measure must be used under such an assumption that every two attributes are independent of each other, which leads to that the similarity calculation between related attributes lacks theory basis. Therefore, building a good case base and carrying out case-based reasoning with related attributes are of great significance.In order to solve these problems, the research introduced in focuses on exploring the efficient method of building case base and the accurate method of case retrieval. For the case base building problem, this paper proposes a FCM-neural network solution for the case-based reasoning together with a FCM-neural network model. First, the original case base is preprocessed by using FCM, and the cases clustered correctly are stored in the new case base and the cases clustered wrongly are ridded.Therefore, the original case base is simplified and the efficient case base is gained. Because the cases are reduced in case base, not only the storage space is saved but also the efficiency of case retrieval is improved. After the simplified case base is gained, the weight of every attribute is determined using a neural network. Therefore, the accuracy of case retrieval yields a certain degree of assurance. For the case retrieval problem, the exsiting self-learning distance measure is improved and a novel method of case retrieval based on particle swarm opitimization (PSO) is proposed. First the PSO is introduced into the self-learning distance measure, to set a new objective function, the new distance measure is gained from a optimization process. Finally, the new distance measure is used in the case retrieval. The self-learning distance measure based on PSO doesn't need to be applied under the assumption that every two attributes are independent of each other, and it takes the correlation between attribute into account. Therefore, it is more reasonable, to carry out case-based reasoning with related attributes which provide the theory basis for the case-based reasoning with related attributes. Finally, this paper applies the proposed method to a complex and irregular problem-determining of the flood-drought situation, which can be a reference for the disaster assessment, and post-disaster reconstruction. The research in this paper considers UCI and flood-rain data as research objects which are applied to realize the simulation analysis, and compares the experimental results of the methods.