Study On Optimization Of Association Rule Algorithm For Customer Segmentation

With the continuous progress and development of Internet technology,the competition in the retail industry has become increasingly fierce.How to make retail enterprises fully play the role of informatization and tap the potential value information of retail enterprises to guide them to maximize profits has important research significance and application value.FPGrowth association rule mining algorithm can mine the relationship between transaction items,but the traditional association rule mining algorithm only mines the relationship between commodity transaction items,lacking the correlation analysis between commodities and customer groups,and can not accurately understand customer needs.Aiming at the above problems,this paper adopts the association rule algorithm of customer-oriented segmentation.To address the problem of poor segmentation accuracy in the RFM customer segmentation model,we consider using a clustering algorithm to improve the quality of customer segmentation.And to maximize enterprise revenue,an improved customer segmentation method is proposed in this paper.The method improves the RFM model by introducing profit attributes,then determines the weights of each index of the improved RFM model by using the entropy weight method,and finally uses the contour coefficient and elbow method to determine the optimal K value before clustering using the K-means algorithm to establish the customer segmentation model and verify the effectiveness of the algorithm.To address the situation that the FP-Growth association rule mining algorithm needs to scan the dataset twice,resulting in low efficiency,this paper adopts adding a hash table and ordered chain table before the item header table of FP-tree to improve the mining efficiency of the FP-Growth algorithm by avoiding repeatedly scanning the dataset,which greatly reduces the search time.Since in the real association rule mining process,the traditional association rule algorithm ignores the semantic metric between transaction items,the semantic metric of the item set is usually characterized by the utility value of the association of transaction items.This results in the phenomenon that there may be a large number of items purchased with high support but low profit,so when the traditional association rule algorithm is used to mine frequent itemsets for the set of goods sold,the utility value of the association rules with high support is not high.To address the above problem of the high-frequency and low utility of goods,this paper improves the FP-Growth algorithm,that is,the HFP-Growth algorithm(High-utility Frequent Pattern Growth Algorithm).The algorithm introduces a utility function and uses the supportutility value framework to generate frequent item sets and then generates efficient and effective association rules.Experimental results show that the HFP-Growth algorithm achieves higher gains.To address the performance degradation of the algorithm for large data sets,this paper parallelizes the association rule algorithm on the Hadoop platform and finally demonstrates the effectiveness and efficiency of the algorithm through experiments.