Research On A Text Association Rules Mining Method Based On Concept Algebra

As the rapid development of Internet and mobile Internet, vast amounts ofinformation from different terminals have been uploaded to the Internet, and theInternet offers us more opportunities to mine resources and information as well.For the user, however, how to get and use the knowledge of what they need fromthe vast ocean of information. It has become an increasingly serious problem.Association rule mining technology as one of the main tasks and methods ofweb text mining technology can help finding some potential repetitive patternsand relevant information. The discovery of the association rules can help us tofind more potential knowledge as well as the subsequent analysis of the textrelevance. The current text association rule mining technology is mainly basedon keywords or needed domain experts to build ontology, so it’s hard to haveenough semantic and deep knowledge mining and the whole process of miningcompletely done by machines and is not confined by the field of text in bothways.In this paper, we use the method of concept algebra’s characteristics of thetext with semantic information, can reflect the text concept relations and canautomatically tectonic characteristics in plain text, with the help of this toolaccording to the text, puts forward the text association rules mining method byconcepts’ indirect and direct relationship from input and output relationship toimprove the traditional association rule generation algorithm, achieve the goal ofeffectively mining concept association rules in text, and through the strength ofassociation rules to quantify the concept of correlation. In contrast of NApriorialgorithm based on keywords relying on support and confidence to generate thefinal rules, this paper use2000economy area texts to mine rules with bothmethods respectively and sort the rules with their strength and support. The method in this paper can better filter some useless rules and find out effectiverules which traditional methods ignored. It proves the rationality and validity ofthe algorithm in this paper.By combining concepts algebraic representation of the text and theassociation rules mined, we put forward the text association degree algorithmbased on the concepts association rules. In text knowledge recommendationsystem, it can recommend texts according to directed association degree. In2000economy area articles, we randomly take3/4texts for association rulesmining, and the remaining1/4texts to calculate the association degree betweentexts to recommend according to the algorithm proposed in this paper. And wetake traditional method based on keywords’ similarity algorithm of the text inthe same texts set, and compare two kinds of results in top10recommendedtext, found that traditional methods can only find similar text content, and thealgorithm in this paper can find articles of text knowledge recommended to theuser which have precise association relationship but contents repeat in lowdegree. Our research tries a new way in complete content based text knowledgerecommendation system, and research results can be applied to complexrecommendation system of content analysis and filter, or used forcategorization/clustering problems based on correlation of texts. It provides anew train of thought and ways for web text analysis and processing.