Research Of Attribute Reduction And Rule Production In Data Mining

As a key data analysis method in Data Preprocess, and based on analysising data without extra information about the data, Rough Set theory has been widely used in many fields of Data Mining, and based on Rough Set, some achivements have arrived. Searching for effective algorithms in Attribute Reduction and Rules production is one of most popular research.Although the most algorithms can get relatively good reduction results, Core attributes should be found in these algorithms which cost the time complexity, and these algorithms ignore the redundant attributes in the selected attributes. Problems will be encountered in dealing with multidimensional data.Based on information theory and Rough Set theory, this dissertation proposes an algorithm, which has less time complexity and deletes the irrelevant attributes. Through test data, this algorithm is proved more effectively in reducing the number of condition attributes in time than other algorithms.A rule production algorithm is proposed to improve the qulitiy of algorithm based on projection and it can solve the noise rules in rules production. In the model of knowledge mining system proposed in this dissertation, new algorithms of Attribute Reduction and rules production are joined together. According to experiments analysis, the better capability of the two algorithms is validated. A database about the scores of the employees in a company is analyzed by the model and the model produces many constructive rules. Many suggestions on the recruit of the company come up according to the analysis of the rules.