XML Document Classification by Frequent Itemset Mining on Menonym Tree

Abstract

XML has been used as a universal format to design the documents on web, because Mark-up language created using XML for any application does not place any restriction on the number of tags that can be defined. The flexibility to create user-defined tags in XML enables smart searches in large data. The structure of XML provides sophisticated proximity measures as the distance between the last word in one element and the first word in the next element is greater than the distance between adjacent words in the same element, even though their physical proximity in the document is similar. As large number of XML documents are used in the web, classification is needed for efficient data retrieval. This paper is based on document classification for XML semi structured data. In this work, it is proposed to exploit the structure feature of XML data and construct a weighted term frequency feature vector with frequent itemset mining of metonymy tree. Two different classifiers: Naive Bayesian Classifier and K-Nearest Neighbourhood Classifier are used for classifying the extracted features. Reuter’s dataset is used for evaluating the performance of classifiers is compared