Based On User Demand Depth-Driven Personalized Recommendation Algorithm

With the advent of the era of the age of the Internet especially the mobile Internet,we are all in an environment of information explosion.Because of the increasing demand for personalized information,users need to filter out a large amount of invalid information.However,there is still no fundamental solution to solve the problem of excessive information.So personalized recommendation algorithm came into being.Personalized recommendation system has received extensive attention since it appeared,and many experts and scholars have proposed their own research methods about personalized recommendation system.The current mainstream recommendation algorithms are content-based recommendation algorithm,graph based recommendation algorithm,collaborative recommendation algorithm and hybrid recommendation algorithm.However,due to over reliance on explicit scoring of data,the current recommendation algorithm suffers from cold boot,data sparseness and recommendation delay,which affects the accuracy of recommendation.This paper studied the optimization problem of personalized recommendation algorithm,focusing on how to make full use of the user's implicit behavior and domain knowledge industry,in order to provide more accurate recommendation results for users.In this paper,a personalized recommendation algorithm based on user needs depth driving is proposed.The algorithm mainly aims at the current recommendation algorithm's cold start,data sparseness,recommendation delay and other issues,and puts forward its own improvement program.In the process of user clustering,the user's implicit behavior analysis is added,and the user's implicit behavior and user's attribute behavior are used to cluster the user.At the same time in generating the recommended list for the user,to join the field of knowledg e,to build industry chain according to the data,the longitudinal horizontal recommend similar products and recommend related products recommended to the user to achieve guide consumption,to help users identify potential demand,produce considerable econ omic and social benefits.Finally,the recommendation system design for closed-loop control system,because the demand will often change,so the system will generate a list of recommended recommendation accuracy at a specific time window detection,can facilitate the timely detection of the accuracy of the recommendation timely adjust the recommendation list.In this paper,the experimental method is used to hold the taobao.com Tianchi competition data,through the data sets were compared to verify the algo rithm,the proposed algorithm and user clustering algorithm and the previous two parts graph recommendation algorithm are compared,the accuracy of the experiment proves that the proposed algorithm significantly improved in clustering accuracy and the accu racy of recommendation.And through the closed-loop design of the recommendation system,the stability of the recommendation accuracy is effectively guaranteed,and the recommended lag is avoided.