Research Of Vertical Search Recommendation System Based On Elasticsearch

In recent years,the Internet in the information age has developed rapidly and has become the main source for people to obtain information.There are two ways of information acquisition:active search and passive recommendation.The information that users need is often highly related to specific vertical areas,such as e-commerce,video streaming,news,etc.The data in these vertical fields is highly structured and closely related to the scene,and is not suitable for the sorting criteria of general search engines.Therefore,traditional general search engines cannot meet the requirements of search and recommendation in vertical fields,and it is imperative to develop personalized search engines based on vertical fields.The paper designs a vertical search recommendation system based on Elasticsearch,which can provide personalized search and recommendation services in the field of e-commerce.The thesis is oriented to the field of e-commerce,based on the Elasticsearch search engine,researches search results ranking correlation technology,personalized recommendation technology,designs and implements a vertical search recommendation system based on Elasticsearch.The main research contents of the thesis are as follows:(1)Design and implement a vertical search recommendation system based on Elasticsearch.As a distributed search framework,Elasticsearch indexes product information and provides retrieval services to the outside world.(2)Adopt a personalized strategy,comprehensively consider the user's historical behavior,the statistical characteristics of the product,and the "query-product" relevance,and reshape the search relevance of the search engine for a specific vertical field,thereby improving the personalized search capabilities of search engines.(3)Study the application of traditional machine learning algorithms in the recommendation field,explore the hybrid strategies of different algorithms,adopt a two-step hybrid recommendation model including recalling and sorting.Use a collaborative filtering algorithm to filter out candidate sets,then use a logistic regression model in sorting stage to sort the candidate sets,avoiding the shortcomings of using a single model.(4)Finally,perform functional and non-functional tests on the system.The experimental results show that the system can meet the needs of personalized search and can complete personalized recommendations for users.