engleski

Classification of textual documents represented in lower dimensional space by supervised Reduced k-means method

The aim of our research is to perform reparameterization of original representations of textual documents in bag of words model in order to enhance performance of their automatic classification. We have used a variant of the Reduced k-means method for a supervised setting which aims at reduction of variables by capturing the clustering structure present in the data. Experiments on Reuters21578 data set show that by representation of documents by this method the same level of precision of classification can be achieved on the higher level of recall as the representation by the benchmark method of the latent semantic indexing.

classification, Reduced k-means method, reduction of dimension, textual documents

nije evidentirano