You are on page 1of 2

Annotated Bibliography Created by Dien Tuan Le (dtl844) W. G. Yi, Lu, M.Y, Liu, Z.

, "Regression Analysis of the Number of Association R ules," International Journal of Automation and Computing, vol. 8, pp. 78-82, 201 1. The authors of this paper studied regression analysis of association rules. The authors argued that if dataset is large, it is difficult to effectively extract the useful rules when the measures of support and confidence were randomly chose n as well as chosen by experience. To address this difficulty, the authors analy zed the meanings of the support and confidence and designed a variety of equatio ns for presenting the relationship between the number of rules and the parameter s. These equations were built by using regression models. Datasets utilized to these equations were collected from the University of Irvine (UCI) machine learn ing repository, which are nursery and letter. Furthermore, the authors utilized statistical approaches including multiple correlation coefficient and hypothesis test to evaluate whether these equations were satisfied and the result showed b etter performance of these equations. This article is not a profound study for association rule mining although the au thors presented a theory of how to design a variety of equations. They only show ed the preliminary state and did not provide the way about how to use these equa tions in different domains. Furthermore, they did not use rigorous statistical t ests to access the quality of rules and did not carry out data mining to find ou t detail association rules for supporting decision. This work is very relevant to my research area as this work provides not only ho w to build a variety of equations for association rule mining, but also how to a pply them to real datasets. However, these equations are only applied to a speci fic dataset of University of Irvine (UCI) while my approach is applied to genera l datasets. L. Shi, X. Jing, X. Xie, J. Yan, "Association Rules Applied to Credit Evaluation ," In 2010 International Conference on Computational Intelligence and Software E ngineering (CiSE), pp. 1-4, 2010. The authors of this paper utilized association rules technology in data mining t o apply to credit evaluation by building a reasonable credit evaluation index sy stem. In the credit evaluation index system, the authors built formula of weight of each index and indicated that the score of each attribute value could be acc urately calculated. In previous studies, the weights of the index were chosen by expert s experience. So, it is too subjective because the weight of the index sho uld be different in the different areas or in a different social economic situat ion. The authors presented the whole process of data mining for credit evaluatio n by dividing into three main stages: (1) data preparation; (2) data mining; (3) expression and interpretation of results. Data preparation phase was necessary for the whole process of data mining because data utilized in data mining were u sually incomplete, messy and repeatable. In a data mining phrase, the authors pr esented some necessary steps in the process of data mining including setting min imum support and minimum confidence; calculating one-dimensional frequent itemse ts; generating possible candidate itemsets for two-dimensional itemsets; countin g two-dimensional candidate itemsets. In last phrase, the authors presented expr ession and interpretation through calculating the weight of very indicator such as age, years of education, and so on. Based on the weight of every indicator, t he authors found out the degree of the influence of the different indicators for returning the entire loan on due. According to theory, this paper is quite detailed in describing the whole proces s of data mining for credit evaluation. However, the details about how to calcul ate the weight of indicator are lacking. For example, there are no details about formula of standard of deviation of confidence. In addition, there is a problem with the built formulas because these formulas are not proven by mathematics. This paper is relevant to my study, which involves applying association rules mi ning to e-commerce domain. However, it does not use the same code of practice as

my proposal research. The evaluation index system will not be used in my resear ch. It will be replaced by a variety of equations.

You might also like