A Rough Clustering Algorithm for Mining Outliers in Categorical Data
Outlier detection is an important data mining task with applications in various domains. Mining of outliers in data has to deal with uncertainty regarding the membership of such outlier objects to one of the normal groups (classes) of objects. In this context, a soft computing approach based on rough sets happens to be a better choice to handle such mining tasks. Motivated by this requirement, a novel rough clustering algorithm is proposed here by modifying the basic k-modes algorithm to incorporate the lower and upper approximation properties of rough sets. The proposed algorithm includes the necessary computational steps required for determining the object assignment to various clusters and the modified centroid (mode) computation on categorical data. An experimental evaluation of the proposed rough k-modes algorithm is also presented here to demonstrate its performance in detecting outliers using various benchmark categorical data sets.
KeywordsData mining Soft computing Rough sets Outlier detection Data Clustering
- 1.Albanese, A., Pal, S.K., Petrosino, A.: Rough sets, kernel set and spatio-temporal outlier detection. IEEE Trans. on Knowledge and Data Engineering (2012) (online)Google Scholar
- 2.Asuncion, A., Newman, D.J.: UCI machine learning repository (2007), http://archive.ics.uci.edu/ml
- 4.Huang, Z.: A fast clustering algorithm to cluster very large categorical data sets in data mining. In: SIGMOD DMKD Workshop, pp. 1–8 (1997)Google Scholar
- 9.Suri, N.N.R.R., Murty, M.N., Athithan, G.: Data mining techniques for outlier detection. In: Visual Analytics and Interactive Technologies: Data, Text and Web Mining Applications, ch. 2, pp. 22–38. IGI Global, New York (2011)Google Scholar
- 10.Suri, N.N.R.R., Murty, M.N., Athithan, G.: An algorithm for mining outliers in categorical data through ranking. In: IEEE HIS, Pune, India, pp. 247–252 (2012)Google Scholar