基于用户评论的混合过滤排序学习推荐系统
首发时间:2018-01-26
摘要:时至今日在大数据环境下,推荐系统扮演者越来越重要的角色。如何从琳琅满目的商品中发掘出用户喜欢的商品成为当下推荐场景中十分重要的任务。本文提出了一种基于用户评论的混合过滤饿排序学习推荐系统。即将用户对电影的评论信息作为特征加入到推荐算法中。并且同时采用基于内容和基于协同过滤的推荐策略,同时提出了一种热度计算的方法。同时将排序学习算法BPR融入推荐算法中,使得最终的推荐列表表现更好。电影推荐系统需要高实时性和有效性。数据库层面电影的属性和相关数据的存储,用户属性和行为数据的存储。考虑到该系统为在线实时系统,所以整个系统的架构分为在线和离线计算两个部分。学习模型的离线计算,有利于系统效率的提升。模型处理部分主要根据推荐方法对数据进行处理得到模型。
For information in English, please click here
Hybrid Filtering Learning to Rank Recommendation System based on user reviews
Abstract:Today in the big data environment, the recommended system to play an increasingly important role. How to find out from the dazzling array of products that users like is the most important task in the current recommended scenarios. In this paper, we propose a hybrid filtering hunger ranking learning recommendation system based on user comments. The user\'s comment on the movie is added as a feature to the recommendation algorithm. At the same time, a recommendation strategy based on content and collaborative filtering is proposed, and a heat calculation method is proposed. At the same time, the ranking learning algorithm BPR into the recommended algorithm, making the final recommendatiHybrid Filtering Learning to Rank Recommendation System based on user reviewson list performed better. Movie recommendation system needs high real-time and effectiveness. Database-level movie attributes and related data storage, user attributes and behavior data storage. Considering that the system is an online real-time system, the architecture of the whole system is divided into two parts, online and offline calculation. Offline calculation of learning model is conducive to the improvement of system efficiency. The model processing part mainly processes the data according to the recommended method to get the model.
Keywords: Artificial Intelligence Recommendation System Hybrid Filtering Topic Model Heat Calculation
基金:
引用
No.****
动态公开评议
共计0人参与
勘误表
基于用户评论的混合过滤排序学习推荐系统
评论
全部评论0/1000