Research And Application Of Improved Apriori Algorithm To Marketing Management System

Classification technology and knowledge discovery based on association rules are two promising research branches of data mining technology. Both have made considerable research values. In this paper, associative classification, as a new topic in data mining, is organically combined by classification rule and association rule together, thereby constructing associative classification system. The associative classification algorithm proposed in this paper first uses a classifier to extract classification rules from training sets, after which they are applied to an improved algorithm of generating association rules to obtain recommended results.This paper briefly describes the theoretical foundation of today’s e-commerce marketing recommendation system. These are data mining, information retrieval and filtering technology. On the basis of this, the article focuses on the classic Apriori algorithm, summarizes the main disadvantages of it and then proposes feasible solutions to resolve the demerits. First, the introduction of the reference database reduces the number of tuples in database when calculating the support count of candidate itemsets, which improves the efficiency of Apriori algorithm to generate frequent itemsets; Secondly, in order to narrow the scope of frequent itemsets when generating strong association rules, the concepts of compacted rule set and improved pruning strategies based on priority as well as an optimized method of generating Apriori strong association rules are introduced; Thirdly, improved ID3algorithm used as prerequisite of generating association rules is associated with the improved Apriori algorithm, which could guide association rules mining and predict consumers’future purchasing behaviors based on which products customers have bought. The improved ID3algorithm in this paper is based on the following three points, namely simplifying the calculation formula of information gain, introducing a penalty function to solve the multi-value bias problem of the original ID3algorithm and discretizing continuous non-nouns attributes.Finally, the improved Apriori algorithm is applied to’ALL E-Commerce Marketing Recommendation Site’developed by the author of this paper. The improved Apriori algorithm firstly divides consumers into different consumption groups and predicts the possibility of purchasing a kind of product in accordance with different types of consumers. Secondly, according to merchandising records of the’ALL E-Commerce Marketing Recommendation Site’, the improved algorithm is utilized to mine potential strong association rules which show the purchasing relationship of different consumer goods, and then to recommend other products to one consumer group who has the tendency to buy a kind of product. The experimental result shows that the improved Apriori algorithm is effective. It improves the quality of strong association rules, effectively reduces the calculation time and because the improved Apriori algorithm is based on the improved ID3algorithm, it could be more targeted to specific customer groups when recommending products.