Feature selection practice for unsupervised learning of credit card fraud detection

Unsupervised Learning processes the massive data and discover the underlying patterns, even though explicit target values are nonexistent. Achieving high predictability for Unsupervised Learning, we practiced to select most influenced feature related to fraud detect system among numerous data. Financial transactions are provided through various channels. On this account, selection of new feature brings increment either on time and cost. In this paper, we practiced the various Feature Selection to detect abnormal transactions exploiting Unsupervised Learning. Here, we select proper features by valuing weight on various Feature Selection Algorithms. The efficiency and accuracy of Feature Selection we practiced are demonstrated by credit card data set. Therefore, it provides rapid response in compliance with feature variance and guide to efficient feature selection.