Weighted Frequent Itemsets Mining Algorithm Based On Matrix Compression And Time Decay

Weighted association rule mining is a technique for mining association rules by giving items weights.The weights can indicate the degree of interest of the user and the importance of items in the data set.Weighted association rules are more in line with people's needs than traditional association rules.But the mining of weighted association rules also has some limitations.For example,the level of user interest may have declined over time,resulting in outdated weighted association rules that are not instructive for current users.Compared with outdated weighted association rules,the recently discovered weighted association rules are what users really need.In the mining process of weighted association rules,the most important step is to mine weighted frequent itemsets.The RWFIM-PE algorithm is a weighted frequent itemset mining algorithm with time decay constraints.The recency calculated by the time decay factor is used to mark the distance between the time point at which the transaction occurs and the end point of the data set.The items included in these transactions are not only affected by the weight,but also by the transaction's Recency value.This makes the last mining result is a weighted frequent itemset that contains the recentness attribute,namely the Recency Weighted Frequent Itemset(RWFI).Association rules generated with recent weighted frequent item sets are more instructive to users in terms of time.But the RWFIM-PE algorithm has some shortcomings.The algorithm follows the idea of the Apriori algorithm,and requires a large number of connection operations when generating(k + 1)item sets.The algorithm has a large execution time overhead and takes up a large amount of memory.In order to solve the problems of the RWFIM-PE algorithm,this paper proposes two weighted frequent itemsets mining algorithm based on matrix compression and time decay(RWFIM-M).This paper mainly completed the following three aspects of work:(1)This paper proposes a algorithm for mining recency weighted frequent itemsets based on matrix compression(RWFIM-M).The RWFIM-M algorithm uses the idea of matrix operations to optimize the process of iterative search candidate sets;the use of matrix compression reduces the search space of the transaction data set,optimizes the process of searching for RWFI,and improves the efficiency of mining RWFI.(2)This paper proposes a algorithm for mining recency weighted frequent itemsets based on projection matrix(RWFIM-PM).Copy the columns of the items contained in the 2-itemset of recent transaction weights with the same prefix into a matrix to form a projection matrix,further compress the matrix,compress the search space of the data set,and further improve the efficiency of mining RWFI.(3)For the algorithm proposed in this paper,three data sets are used to conduct comparative experiments from multiple angles.Analyze the experimental results and prove that the proposed RWFIM-M algorithm and RWFIM-PM algorithm have good performance.