Intelligent E-commerce Recommendation System Based On Big Data Of Communication Operator

With the rapid development and improvement of modern Internet technology,as well as the increasing stability and faster speed of residential broadband,more and more users are using various electronic terminals,such as computers,Tablet,TV box and especially mobile phones,to obtain rich Internet content from residential broadband.Based on the Internet technology and information network technology,e-commerce shows tremendous potential for development,and especially online shopping has been greatly appreciated by users.Nowadays,big data is developing rapidly.By collecting residential broadband users' online shopping data and behaviors on the e-commerce platforms,big data provides great opportunities for users' data mining,advertisement delivery and product recommendation in the future e-commerce platform enterprises,but also provides long-term development and utilization value for the future big data development.On the basis of having unique advantages of more complete users' online shopping traffic compared with other e-commerce platforms,personalized recommendation of online shopping for residential broadband users not only provides a reliable way for communication operators to cash the flow,but also provides high-quality big data services for residential broadband users.Personalized recommendation is of far-reaching significance and value in the field of e-commerce.Accordingly,this paper proposes a smart e-commerce recommendation system based on big data of communication operators' pipelines.Since most e-commerce sites are not fully encrypted,the recommendation system uses DPI technology to collect data from non-encrypted data packets and perceive user identity information and user behaviors.At the same time,the recommendation system uses the pre-established universal tag library to tag the user's behavior.Finally,the recommendation system uses the mixed recommendation algorithm to predict the users' interests based on the tags extracted by users.For this recommendation system,the main innovations of this paper are as follows:1)The signature keywords automatic extraction technology is used to mine the signature keywords of the HTTP traffic packets,and the signature keywords are denoised through the big data platform to improve the accuracy of signature keywords,replacing the traditional way of off-line naked eye comparison.2)The relevance of the terms is calculated based on the improved TF-IDF algorithm.The feature vectors are clustered by K-means,and the universal tag library is formed based on the cosine similarity between the vectors.3)According to the user's previous interest products and the extracted product classification tags,this paper use the weighted fusion collaborative filtering algorithm based on improved SVD and the improved GBDT algorithm based on users' interest lasting value and the popularity of date to generate recommendation for users.