Published Paper Details:

In data mining, the filtering feature selection approach is a well-known feature selection strategy. When the dimension of features is big, the filtering approach outperforms other feature selection methods in many circumstances. Because there are so many filtering methods described in prior work, how to choose an acceptable filtering method for a particular text data collection has become a "selection problem." Because determining the appropriate filtering method is frequently intractable in practise, this research proposes a different approach. We present a framework for feature selection that combines the results of several filtering approaches. In fact, rank aggregation, or the process of creating a superior rank list from many rank lists, is a hot issue being researched across many fields. In this research, we describe FR-Borda, a new feature selection approach based on the presented framework and Borda rank aggregation techniques. The novel feature selection approaches obtain better or equivalent results to conventional filtering methods in an empirical evaluation using tweeter text data, demonstrating the usefulness of our system.