Interestingness of measures: A statistical prospective

Author

Selvarangam, K. ; Ramesh Kumar, K.

Author_Institution

Dept. of Comput. Sci. & Eng., Hindustan Univ., Chennai, India

fYear

2014

fDate

27-29 Nov. 2014

Firstpage

209

Lastpage

213

Abstract

Ranking interestingness measure is a necessary part in the process of knowledge discovery from the extracted rules. Since the range of values of Interestingness measures are not unique, identifying a perfect measure is a challenging question to data mining community. Homogeneity of a measure always varies from 0 to 1; hence we ranked the measures by calculating the homogeneity coefficient, using the score of the respective measure on a set of rules. Also we introduced heuristic association measures, U Cost, S Cost, R Cost, T Combined Cost and ranked with existing measures using the ranking algorithm. Our measures are placed in better position on ranking, compared with the existing measures.

Keywords

data mining; statistical analysis; R-cost; S-cost; T-combined cost; U-cost; data mining community; heuristic association measures; interestingness measure ranking; knowledge discovery; measure homogeneity coefficient; ranking algorithm; rule extraction; rule set measure score; statistical analysis; Association rules; Atmospheric measurements; Educational institutions; Knowledge discovery; Particle measurements; Probability; Association rules; Data mining; Homogenity coefficient; Interestingness meausures; Ranking; Variability coefficient;

fLanguage

English

Publisher

ieee

Conference_Titel

Contemporary Computing and Informatics (IC3I), 2014 International Conference on

Conference_Location

Mysore

Type

conf

DOI

10.1109/IC3I.2014.7019800

Filename

7019800