| With the development and increasing application of Case-Based Reasoning (CBR), the scale of the case base is expanding continuously, which brings two consequences. On one hand, the more cases the case base contains, the more knowledge it can provide and the more capable of the CBR system. On the other hand, the increasing of the inconsistent, incomplete, redundant and useless cases which came along with the expanding scale lowers the quality of the case base and impairs its performance. Therefore, the method of control the case base scale, maintaining the problem-solving ability of CBR system and fastering the accessing of the case base has become the focus of the research and application of CBR.Multiple strategies are used in this thesis to maintain the case base and improve the original strategy, so that the problem solving ability of CBR system is obtained, and the scale of the case base is properly controlled at the same time.Firstly, the original case base is clustered using Adaptive fuzzy c-means algorithm, the correctly clustered cases are stored in the case base according to their type logically. The incorrectly clustered cased are deleted to remove the redundancy, thus the original case base is simplified, its storage structure is optimized and a more efficient case base is obtained. The reduction of the case number not only saves the storing space of the case base, but also increases the case searching efficiency.Then the two level index mechanism constructed according to the clusters and case features, the cache case base and the real-time case storage strategy enable the case base to add new case and update the index rationally.Finally, the case features are reduced using neighborhood rough set based forward greed feature selecting algorithm, the size of the case base is reduced by deleting the unnecessary features.The performance of the case base is improved effectively through general maintenance and improvement of the original technology. |
PDF Full Text Request