Research And Application Of E-Commerce Advertisement Recommendation Model Based On User Interest

The rapid development of commodity trading websites has prompted the emergence of various e-commerce marketing methods,which online advertising is one of them.However,advertisement click-through rate and conversion rate are lower because of the simple widely distribution so that advertising resources waste at present.Under the impetus of the precision of advertising,network-oriented advertising came into being.The main idea is to mine user implicit information through existing information technology and build corresponding user interest models so as to recommend advertisements with higher degree of match to the model.However,current network-targeted advertising still has problems such as low efficiency,unobvious user interest characteristics,and low matching of advertisements.Therefore,it is necessary to study and design high-efficiency,retain interest characteristics to implement user interest model for accurate advertisement placement.In response to the above issues,the main research contents of this paper are as follows:(1)Analyzed the optimization methods of the existing feature methods,and proposed a feature selection method that has combined the bee colony and gradient elevating decision tree algorithm.In this paper,this method has been used to optimize the user behavior under the complex e-commerce environment,retained the high importance features,whose weights have been formed by their importance.(2)Studied the user's access to web content,and extracted user interest feature words through Chinese word segmentation and stop word filtering technology.We have jointly optimized the user's network-wide behavior features,built a user interest model based on improved vector space model to achieve multiple user interests reflect.(3)As to the semantic mismatch of feature words,a method of semantic extension and similarity calculation based on knowledge graph has been proposed.The method expanded interest feature words on the basis of knowledge graph—DBpedia,and used an improved semantic correlation algorithm based on n-Gram and semantic distance to calculate semantic relevance,and combined the original interest feature weights to obtain extended word weights to further improve the accuracy of user interest.Experiments have showed that the advertisement recommendation model based on user interest had good effect and had certain significance and practical value.