No abstract

No abstract

Text clustering is gaining importance among researchers because of rapid increase in the availability of online text collections without class labels. It helps to organize, summarize and retrieve useful information from corpora. High dimensionality of text datasets leads to poor performance of clustering algorithms. Dimensionality can be reduced using feature extraction or feature selection methods. Feature selection methods scale well and are easy to interpret. An unsupervised univariate filter feature selection method was proposed for dimensionality reduction. The proposed method outperformed nine other filter methods reported in the literature, by identifying most relevant features that lead to good clustering performance on eight popular text datasets.

No abstract