| Summary: | 碩士 === 雲林科技大學 === 資訊管理系碩士班 === 96 === This paper proposes an integrated classification method based on the rough set approach. We use the Cumulative Probability Distribution Approach (CPDA) to discrete the numberic data. From discretization result, use the k-means algorithm to cluster library borrowing habits. The clustering result assign as class attribute of the dataset, extracting classification rules cooperate with characteristic attributes of students. Result can do the reference of active library borrowing service. In empirical case study, we use the data sets drawn from a newly established college in central Taiwan. Combining the library borrowing records in the library information system with the student profiles in the academic affair information system, there are 6,121 students who borrowed books and there are 256,254 Chinese books in the library borrowing history records. Using the proposed approach, we divide the student''s borrowed habits into three clusters, obtain 519 classification rules. The accuracy rate of training data set is 99.9% and the accuracy rate of test data set is 89.4% using the result rule set. Compare with Bayesian classification, C4.5 decision tree and RIPPER, the proposed approach has improved accuracy.
|
|---|
