Research On Personalized Recommendation Method Based On Association Rule Mining

In recent years,with the rapid development of Internet and information technology,the scale of the Internet has been greatly improved.The information it provides to users is also increasing dramatically.We are facing an era of big data.In this mass of data,how to get the most personal resources is a very prominent problem.Existing traditional search engines and portal sites,to a certain extent,alleviate this problem,but they can not fully meet the needs of people.At present,the recommendation system is an effective solution to solve the problem of information overload.But it can't provide very good personalized service.Under this background,recommendation system is an effective solution to the problem of "information overload".It has received extensive attention from the academic and industrial circles,obtained great achievements in practical.The recommended approach of recommendation system include collaborative filtering recommendation,association rule recommendation,content-based recommendation,and context-aware recommendation method.Association rule is the one of widely used recommendation methods in application.With respect to the collaborative filtering recommendation method,it's a direct recommendation method.This method mining potential association relationship from the overall data,has nothing to do with personal preferences and recommend interested information for users.Apriori is a classical algorithm of association rule,the algorithm mainly uses layer by layer iteration to generate high-dimensional frequent itemsets by using low dimension frequent itemsets.At home and abroad,the algorithm is optimized from hash,sampling,mapreduce model and so on.This methods improve efficiency of extracting frequent itemsets mainly from divide-and-rule and data-io.But it lacks consideration of recommendation balance and efficient processing between popular and unpopular data.In this paper the problem of mining frequent itemsets based on association rule is revaluated and analyzed.It is found that in the data pruning process,we can use the statistical support degree to pre prune the candidate itemsets when the frequent itemsets are combined,and combine the Top-N recommendation theory to optimize the algorithm.In order to improve the quality and efficiency of personalized recommendation and balance the recommendation weight of popular and unpopular,a notion of k-pre association rule is defined and improved the evaluation metric called RecNon,the pruning strategy based on k-pre frequent itemset is designed.Moreover,an association rule mining algorithm based on the idea is proposed,which optimized the Apriori algorithm and is suitable for different evaluation criteria,reducing the time complexity of mining frequent itemset.The theoretic analysis and experiment results show that the method improved the efficiency of data mining and achieved higher RecNon,Fmeasure and precision of recommendation,and efficiently balanced the recommendation weight of popular data and unpopular one.