Research And Application Of Item Correlation Search Algorithm In Information System

The function of information search method is to find the information data that users need and care about quickly and accurately in the information system and improve the efficiency and quality of search.In recent years,the rapid development of Internet technology drives the leap of data processing capability of information system and accelerates the pace of system update and iteration.For example,chunyu doctor,qunar,Tmall and other major information systems provide users with massive data information in order to compete for the number and access flow of users,which enriches users' choices and also causes the phenomenon of "information overload",which interferes with the process of users' information selection and affects the experience of the system.Faced with the problem of "information overload",it is of great value and significance to help users eliminate the interference of irrelevant information and quickly and accurately locate the results of interest.This paper mainly discusses the research and application of information search methods from the following three aspects:The application of information search method first needs to build the user's interest preference model.In this paper,lm-bp neural network algorithm was used to train the user preference model.First,the characteristic attribute matrix was established according to the characteristics of users and projects,and the matrix was dimensionally reduced.Then,lm-bp algorithm was used for training to construct the user-project scoring matrix and predict the projects without scoring.Complete the establishment of user preference model.Through experiments,it can accurately reflect the user's interest and preference.Secondly,the selection and implementation of search recommendation algorithm.The traditional collaborative filtering algorithm has three problems: sparsity,"cold start" and scalability.(1)in the face of sparsity problems,this paper USES the non-target user type discrimination theory to judge the recommendation ability of users.The unscored value filling method based on the nearest neighbor theory is described to alleviate the sparsity problem.(2)in the face of the "cold start" problem is collected through the web log network access sequence,and explained by computing the similarity of the network access sequence,in order to search for the nearest neighbor set of the new user method.(3)facing the scalability problem,this paper describes a collaborative filtering incremental update mechanism that ADAPTS to the change of user interest,and this mechanism has good performance.(4)finally,this paper USES the collaborative filtering algorithm based on gray association clustering to calculate the similarity between users and obtain the preference model of new users by weighted average.Finally,the search method is validated by building a search platform.Although this paper focuses on information system,limited by the research time and my own level,only e-commerce system,a typical information system,is selected as the background for discussion and verification.This subject is supported by the experimental data and experimental conditions of the full set network e-commerce platform.