Research On Personal Recommendation Approach Based On Multi-Objective Evolution Algorithm

With the rapid development of information network technology,the e-commerce system has penetrated into every aspect of life.Through a variety of online sales platforms,people can complete the shopping and trading of goods without leaving their homes,which saves a lot of purchase time.However,the ever-expanding e-commerce system has also caused some troubles.For example,in the face of massive product information,users need to spend a lot of time finding the part they interested.As a technology based on information collection and knowledge discovery,recommendation systems are increasingly being used in e-commerce systems.It should be noted that as people's needs become more diverse,it is not enough to simply pursue the accuracy of recommendations.In order to meet the individual needs of users,on the basis of traditional recommendation techniques,it is necessary to add other performance indicators,such as diversity and novelty.It has been found that the conflict relationship between accuracy,diversity and novelty make these three goals optimal at the same time.Therefore,it is difficult to make these three goals optimal at the same time.How to get a recommendation that performs well on the above indicators has become a difficult problem to be solved.In the process of solving practical problems,we often need to consider a variety of factors to determine the feasibility of the program.This is the so-called multi-objective optimization problem,which means to obtain a set of optimal values in the case of conflicts between objectives.Based on the conflict relationship between the objectives of the current personalized recommendation system,some scholars have proposed to convert the personalized recommendation problem into a multi-objective optimization problem,and optimize multiple recommendation indicators at the same time.At present,the mainstream personalized recommendation algorithm based on multi-objective optimization mainly takes accuracy and diversity as dual targets.Although it can provide users with more diverse products than traditional algorithms,the popularity of this part of products is also high.Obviously,this multi-objective algorithm that recommends popular goods to users is not conducive to mine unpopular products.In the long run,it is not helpful to improve the purchase rate of the system.Based on the above theoretical analysis,this paper has carried out the following in-depth research:(1)An improved personalized recommendation algorithm based on MOEA-ProbS(MOEA-PGMA)is proposed.Based on the accuracy and diversity as the evaluation indicators,the novelty is introduced into the personalized recommendation,and the three-objective personalized recommendation model is innovatively proposed to improve the ability of mining users' potential interests and unpopular products;An adaptive mutation method suitable for multi-objectives is proposed on the three-target recommendation model(MOAM).Due to the problem of premature convergence occurs in the process of population evolution of MOEA-ProbS,MOAM is applied to improve the search ability of the population and increase the accumulation of excellent genes to obtain better individual populations than the MOEA-ProbS algorithm.The experimental results show that the MOEA-PGMA algorithm can improve the solution of the above problems.(3)A concurrent implementation method of multi-objective recommendation algorithm(MOEA-PGMA-MC)is researched,which uses multi-threaded concurrency technology to improve the execution efficiency of the algorithm.While verifying the effectiveness of the algorithm in the previous chapter,it is found that as the number of users and targets increases,the computational scale of the algorithm will be further expanded,which makes the program run slowly and the execution efficiency is low.Through the analysis of the evolution process,it is found that to use multi-threaded concurrency in the algorithm is feasible to improve the execution efficiency of the algorithm.The experimental results show that the efficiency of the algorithm is significantly improved,compared to the execution time in the serial environment,on the 1000-dimensional large-scale recommendation problem.In general,this paper first systematically studies the issue of personalized recommendation.Firstly,aiming at the lack of unpopular data and user preferences in the current recommendation system,a three-objective recommendation model is proposed innovatively;the recommendation algorithm based on the model is designed and optimized;finally,the implementation performance is introduced by applying multi-thread technology.The experimental results show that the multi-objective personalized recommendation algorithm proposed in this paper can generate multiple sets of recommended schemes with good accuracy,diversity and novelty in one operation and the execution efficiency is obviously improved by introducing multi-threaded technology into the algorithm.