Optimal feature selection for multivalued attributes using transaction weights as utility scale

Abstract

Attribute selection procedure is a key step in the process of Knowledge Discovery in Database (KDD). Majority of the earlier contributions of selection methods can handle easier attribute types. Such methods are not for multivalued attributes that comprise multiple values in simultaneously. Majority of the existing attribute selection methods can manage simple attribute types like the numerical and categorical. The methods cannot fit multivalued attributes, which are attributes that constitute multiple values simultaneously in the dataset for same instance. In this manuscript, a contemporary approach for selecting optimal values for features of multivalued attributes is proposed. In the proposed solution, the method is about adaptation of utility mining based pattern discovery approach. For evaluating the proposed approach, experiments are carried out with multivalued and multiclass datasets that are submitted to k-means clustering technique. The experiments show that the proposed method is optimal to assess the relevance of multivalued attributes toward mining models such as clustering.