The Design Of Product Marketing System Based On Association Rules Algorithm

After years of development,e-commerce has formed a relatively mature method in product recommendation.Generally speaking,there are three types of recommendation,namely intelligent retrieval,user similarity and timed push.however,for products,its recommendation method is too old,having not been properly updated and optimized.At present,the recommendation is only based on the historical preference of the users,without considering the influence of the users' behavior on the product selection.Users' preferences for products will change according to the changes of users' behaviors,which requires rapid response to the changes and find new matching products to meet users' new needs and preferences.However,at present,there is still a big problem,and the recommendation is obviously lagging behind.At the same time,there is a lack of scoring in collaboration,and is only replaced by the same default value,which also makes the evaluation matrix computation general,not including the users' preferences for product difference,education background and salary to calculation.So it also can't calculate the user's special demand,and such recommendations are not practical.In order to solve the above problems,it is necessary to establish a new method of recommendation,which is based on the dynamic behavior of users and updates the recommendation in real time.Firstly,it is important to monitor the uers' operation throughout the whole process,and collecting and sorting out the uers' new product demand and clicking on other products timely,so as to make the latest judgment on the users' demand and recommend produsts to the user.Only in this way can the recommendation solve the problem of lagging and be timely.In addition,clearly refining the score of collaborative computing of products,and accurately filling the score value of the missing part,so as to make more accurate the calculation of similarity,and then screening the results,making the obtained information of data can be more reliable.Discussing the technologies and theories involved in this paper,such as reconstruction of data mining theory and recommendation algorithm well as working ethic,programming language,Hadoop framework,etc.,to explore the possibility of various theories and technology combination,and lay a foundation for building the marketing system of electronic equipment charging products based on associationrules.Then,based on the shortcomings of the current association rule algorithm in the recommendation system,this paper proposes improvement suggestions from the perspective of FP-growth algorithm and parallel association rule method,and constructs the recommendation model,and then compares and verifies the improved model,to prove the performance superiority of this model.Specific implementation of marketing system based on association rule algorithm.The current data acquisition module can be carried out in two modes,database or message queue,and they are used to store the information of various behaviors of users.According to the logic of recommendation,it can be divided into two categories:real-time recommendation and collaborative recommendation,which also enables the recommendation module to expand from these two sub-modules.Real-time products recommendation collectsand sorts the users' latest behavior to carry on the,using information queue to save the information,when the users have the new operation,update the product requirements,or click on the new product,it can trigger new match with the change messages in the queue,allowing users to get the latest online matching and association recommending products,satisfying the uses' quick purchase intention.In terms of the development of collaborative recommendation of products,it is necessary to predict the missing parts in the scoring matrix to fill in different new values,which can greatly improve the quality of similarity.Then,through calculation and screening of the calculation results,the recommendation can be more in line with users' preferences and enhance the reliability.The design of interaction module on the foreground page is based on the responsive layout,which can also shorten the response time and improve the efficiency.System implementation and testing.The operation of this system is based on the application of Java platform.Storm is used to complete real-time recommendation of products.And collaborative recommendation of products is realized by Mapreduce.The results show that the system is capable of recommending products with a high rate,reliability and stability.In addition,after estimating the missing value of the scoring matrix and carrying out the necessary filling,there is a great improvement in the relevant evaluation indexes recommended by the collaborative recommendation.