Design And Implementation Of E-Commerce Recommendation System Based On Elasticsearch

In today's era of rapid development,with the rapid development of mobile Internet,the huge convenience brought by this is slowly permeating into People's Daily life,covering all aspects of clothing,food,housing and transportation.Both offline O2 O platforms such as Meituan and dianping,or e-commerce platforms such as jingdong and taobao,in which the commodity services and store management are growing exponentially with huge contents and different characteristics.In the face of a huge number of commodities and stores,how to find the most satisfied becomes the key to improve the browsing click through rate and transaction conversion rate of commodities or stores.Therefore,a kind of Elasticsearch based e-commerce recommendation system architecture is proposed by combining the search engine with the recommendation system in the field of e-commerce based on the fast retrieval nature of search engines.In this paper,in the field of electricity was designed and implemented a distributed electrical contractor recommendation system based on Elasticsearch search engine,first of all,this paper introduces the research background and significance,and analyzes the research status of search engine at home and abroad,then introduces the system involved in the process of the implementation of the techniques and algorithms,next elaborated the system functional and nonfunctional requirements,technical scheme,architectural design,technology design and system implementation,finally,the system was tested and performance analysis,through click-through rate forecast model can explain and has the practical significance of the evaluation index,That is,the user's browsing click through rate and transaction conversion rate has a significant increase.In the system search section,this paper introduces the full-text retrieval capability of Elasticsearch search engine,designs the Elasticsearch multi-field query and scoring principle by combining with the TMDB open source data source,and realizes the customizable scoring sorting logic.It can be done via the Chinese word segmentation in Chinese word segmentation,with logstash-input-JDBC build full amount and non real-time incremental index,in the search engine architecture,by customizing tumble word and synonym expansion to rich search accuracy,and through a speech analysis and reshape the correlation algorithm,build an understandable semantic search engine.In the system recommendation part,this paper adopts the keyword extraction algorithm,multi-channel recall recommendation algorithm,and the sorting algorithmbased on the mixed GBDT and LR models to realize personalized recommendation and online click rate estimation for user data.This recommendation system applies to the recommendation system with items and evaluation,such as movie recommendation system,recommendation system,at the same time,malicious evaluation of the effectiveness of the filter need to accumulate a large number of users of the original data,the recommended system is realized in this paper in the user data is less stage,malicious evaluation filter can only rely on public data sets,after the system data accumulated enough,switch to a historical user evaluation.The final realization of this paper is an e-commerce system combining recommendation and search,which combines the user's historical behavior with different recommendation algorithms to achieve recommendation recall and recommendation ranking,and explains the system implementation process by showing the detailed design scheme of each major functional module.In the final system test,the CTR of different recommendation algorithms was estimated and compared.In the test results,it was found that the recall rate and accuracy rate of the system were improved,and the impact on the CTR of browsing and transaction conversion rate were also improved.