Document Grouping by Using Meronyms and Type-2 Fuzzy Association Rule Mining

Abstract

The growth of the number of textual documents in the digital world, especially on the World Wide Web, is incredibly fast. This causes an accumulation of information, so we need efficient organization to manage textual documents. One way to accurately classify documents is using fuzzy association rules. The quality of the document clustering is affected by phase extraction of key terms and type of fuzzy logic system (FLS) used for clustering. The use of meronyms in the extraction of key terms to obtain cluster labels helps obtaining meaningful cluster labels and in addition ambiguities and uncertainties that occur in the rules of type-1 fuzzy logic systems can be overcome by using type-2 fuzzy sets. This study proposes a method of key term extraction based on meronyms with an initialization cluster using fuzzy association rule mining for document clustering. This method consists of four stages, i.e. preprocessing of the document, extraction of key terms with meronyms, extraction of candidate clusters, and cluster tree construction. Testing of this method was done with three different datasets: classic, Reuters, and 20 Newsgroup. Testing was done by comparing the overall F-measure of the method without meronyms and with meronyms. Based on the testing, the method with meronyms in the extraction of keywords produced an overall F-measure of 0.5753 for the classic dataset, 0.3984 for the Reuters dataset, and 0.6285 for the 20 Newsgroup dataset.

Keywords

• association rule
• document clustering
• type-2 fuzzy sets
• meronyms
• overall F-measures
• textual documents.