基于Spark的商品推荐系统的设计与实现

Design and Implementation of Product Recommendation System Based on Spark

本文通过用户对商品的评分信息以及详细信息进行分析建模。基于Spark开源平台,利用SparkMLlib和SparkStreaming,结合交替最小二乘法、余弦相似度、同现相似度等算法,实现了针对不同用户和商品的协同过滤推荐和相似推荐等离线推荐模式,以及对实时日志数据进行处理,形成实时推荐优先级列表的实时个性化推荐。满足用户的商品购买需求,增加了电商平台的销售额,极大程度的降低了信息过载给商家和用户所带来的困扰。

This paper analyzes and models the scoring information and the detailed information of the commodities.Based on Spark open source platform,using SparkMLlib and SparkStreaming,combined with the alternative least square method,cosine similarity,co-occurrence similarity and other algorithms,it realizes the collaborative filtering recommendation and similar recommendation for different users and commodities,and processing the realtime log data to form the realtime personalized recommendation of the realtime recommendation priority list.It can meet the demand of customers,increase the sales of e-business platform,and reduce greatly the problem of information overload to businesses and users.