A Game-Theoretic Rough Set Approach for Handling Missing Data in Clustering
An important issue in clustering is how to deal with data containing missing values. A three-way decision making approach has recently been introduced for this purpose. It includes an added option of deferment which is exercised whenever it is not clear to include or exclude an object from a cluster. A critical issue in the three-way approach is how to decide the thresholds defining the three types of decisions. We examine the role of game-theoretic rough set model (GTRS) to address this issue. The GTRS model induces three-way decisions by implementing a game between multiple cooperative or competitive criteria. In particular, a game in GTRS is proposed which realizes the determination of thresholds from the viewpoint of tradeoff between accuracy and generality of clustering. Experimental results are reported for two datasets from UCI machine learning repository. The comparison of the GTRS results with another three-way model of (1, 0) suggests that the GTRS model significantly improves generality by upto 65% while maintaining similar levels of accuracy. In comparison to the (0.5, 0.5) model, the GTRS improves accuracy by upto 5% at a cost of some decrease in generality.
KeywordsClustering Missing data Three-way Game-theoretic rough sets
This work was partially supported by a Discovery Grant from NSERC Canada and Indigenous Student Scholarship from HEC Pakistan.
- 4.Eye, A.V.: Statistical Methods in Longitudinal Research: Principles and Structuring Change. Statistical Modeling and Decision Science, vol. 1 (2014)Google Scholar
- 10.Lichman, M.: UCI machine learning repository (2013). http://archive.ics.uci.edu/ml. Accessed 9 Feb 2017
- 18.Yu, H., Su, T., Zeng, X.: A three-way decisions clustering algorithm for incomplete data. In: Miao, D., Pedrycz, W., Ślȩzak, D., Peters, G., Hu, Q., Wang, R. (eds.) RSKT 2014. LNCS (LNAI), vol. 8818, pp. 765–776. Springer, Cham (2014). https://doi.org/10.1007/978-3-319-11740-9_70CrossRefGoogle Scholar