CRAFT: ClusteR-specific Assorted Feature selecTion
This work addresses clustering with feature selection for mixed data types, offering a practical tool for data analysis, but it appears incremental as it builds on existing nonparametric MAP-based models.
The paper tackles the problem of clustering with cluster-specific feature selection for assorted data, presenting CRAFT, a framework derived from asymptotic log posterior formulations that does not require specifying the number of clusters a priori and performs favorably on real datasets.
We present a framework for clustering with cluster-specific feature selection. The framework, CRAFT, is derived from asymptotic log posterior formulations of nonparametric MAP-based clustering models. CRAFT handles assorted data, i.e., both numeric and categorical data, and the underlying objective functions are intuitively appealing. The resulting algorithm is simple to implement and scales nicely, requires minimal parameter tuning, obviates the need to specify the number of clusters a priori, and compares favorably with other methods on real datasets.