Machine learning techniques are widely used to extract valuable knowledge from data, but their performance might get affected when the proper features are not selected during the extraction. Feature selection is introduced to search relations and disclose possible redundant or irrelevant features in a case study; this search performed can either be supervised or unsupervised.In the present work, we propose an unsupervised feature selection algorithm using: (1) relative dependency to search similarities between features, (2) a clustering algorithm to group similar features, and (3) procedure to select the most representative feature to obtain a reduced feature space. At present, dealing with data analytics of unlabelled data is quite difficult as the patterns they produce are highly random. By using techniques such as Linear Regression, KNN mean, Decision Tree and Neural networks helps in recognising such patterns. Unsupervised learning methods have to work with the given attributes. Bringing a comparative study of different algorithmic strategies are suitable for observing the patterns.
Volume 11 | 08-Special Issue
Pages: 1252-1257