Research On Personalized Recommendation Based On Fuzzy Concept Lattices

The Formal concept analysis (FCA) was started by Professor Wille, German mathematician,in1982. After three decades of development, the FCA has been applied to many areas, such asknowledge discovery and machine learning and so on. The core data structure of the FCA isconcept lattice, so the core of the application is building the concept lattice. Based on the theresults of the previous, This paper improves an incremental concept lattice constructionalgorithm, and the theoretical analysis and experiments have proved that the improved measuresimprove the speed of the algorithm. This paper also improved the association rule miningalgorithm on the concept lattice, and applied the algorithm into the personalized recommendationsystem. The main contents are summarized as follows:(1) Introduction to the basic knowledge which are used in this paper, includes the datamining, personalized recommendation, the fuzzy concept lattice as well as the association rules.(2) Has improved an incremental concept lattice generation algorithm, and extend it to thebuilding of the fuzzy concept lattices. The original concept lattice generation algorithm traversesthe concept lattice by traversing an index tree, and generates the new concept lattice in theprogress. This paper gives many improvements of traversing the index tree. We found that theindex tree traversal implies an ordering by thinking carefully, and the ordering is the dictionarysequence. This paper further exploits the ordering by sorting the attribute set (the letterscollection) of the new object with ascending, and Setting a ‘unik’ property for each tree node ofthe index tree. The ‘unik’ represents the characteristic of the edge between the current tree nodeand its parent node. And you can convert part of judging the type of the concept lattice node tothe size comparison of letters, thence to reduce the complexity of the algorithm, narrow the rangeof traversing the index tree, reduce redundant collection operations, and optimize the originalalgorithm.(3) Based on the incremental concept lattice generation algorithm, this paper improves theassociation rule mining algorithm on the concept lattice. The main part of the algorithm iscalculating intent reduction of the grid node. This paper focuses on an incremental intentreduction algorithm of the concept lattice, and integrates the algorithm into the previousoptimized incremental concept lattice building algorithm. Because of the improved measures, theincremental concept lattice generation algorithm becomes faster, hence, the new algorithm certainly better than the original one. The theoretical analysis and experiments also proves thesuperiority of the algorithm. Compared to the previous incremental concept lattice generationalgorithm, the result of the intent reduction algorithm is the quantitative concept lattice, whichcontains the intent reduction set, and is suitable for the mining association rules very much and isapplied to the subsequent association rule mining job.(4) Design a personalized recommendation system on the video site, and implement it.