| The aim of data mining is extracting effective, innovative, potentially useful, and ultimately understandable patterns from databases. In recent years, mining association rules has received a great deal of attention, because it can capture the important relationship between the data, and the forms of rules been mined are concise and easier to understand.Firstly, this paper introduces the background, conception and definition, methods of data mining, discusses technology of association rules mining in detail, including its target, classification and algorithms, and especially describes the data processing methods for different attribute types.Although many algorithms have been proposed to find association rules in databases with Boolean attribute and quantitative attribute, few have tried to do so with ordinal attribute. For ordinal attribute, traditional mining algorithms usually handle it as categorical attribute, mapping eachpair to Boolean attribute, then use Boolean attribute mining method. The ordinal attribute value is only used to differentiate between different objects as the nominal, but its order is neglected. This paper presents an advanced method. First of all ordinal attribute is converted into categorical attribute through membership fuction of fuzzy relation, so that all items in transacts is no longer a Boolean value of "1" or "0", but a floating-point number between [0, 1]. Secondly the support of floating-point number is defined to calculate the degree of support. Then we can use an Apriori-like candidate set generation-and-test approach to mine association rules. This advanced method takes into account the impact of the order better, and enhances the validity of patterns found.Finally, the advanced algorithm is applied in college teaching evaluation databases. The results of mining are analyzed to provide more decision support information for improving the quality of teaching. At the same time, the results are used to evaluate our method. |
PDF Full Text Request