基于情感分析和LDA主题模型的协同过滤推荐算法

发布时间：2018-08-01 18:38

【摘要】：协同过滤推荐算法通常基于物品或用户的相似度来实现个性化推荐,但是数据的稀疏性往往导致推荐精度不理想。大多数传统推荐算法仅考虑用户对物品的总体评分,而忽略了评论文本中用户对物品各个属性面的偏好。该文提出一种基于情感分析的推荐算法SACF(reviews sentiment analysis for collaborative filtering),该算法在经典的协同过滤推荐算法的基础上,考虑评论文本对相似度计算的影响。SACF算法利用LDA主题模型挖掘物品潜在的K个属性面,通过用户在各个属性面上的情感偏好计算用户相似度,从而构建推荐模型。基于京东网上评论数据集的实验结果表明,SACF算法不但可以有效地改善传统协同过滤推荐算法中数据稀疏性的问题,而且提高了推荐系统的精度。
[Abstract]:Collaborative filtering recommendation algorithms usually implement personalized recommendation based on the similarity of items or users, but the sparsity of data often leads to unsatisfactory recommendation accuracy. Most of the traditional recommendation algorithms only consider the user's overall rating of the item, but ignore the user's preference for each attribute surface of the item in the comment text. This paper presents an affective analysis-based recommendation algorithm (SACF (reviews sentiment analysis for collaborative filtering), which is based on the classical collaborative filtering recommendation algorithm. Considering the influence of comment text on similarity calculation. SACF algorithm uses LDA topic model to mine the potential K attribute surfaces of items and calculates user similarity through user's emotional preference on each attribute surface and then constructs a recommendation model. The experimental results based on the review data set on JingDong network show that the proposed algorithm can not only effectively improve the data sparsity in the traditional collaborative filtering recommendation algorithm, but also improve the accuracy of the recommendation system.
【作者单位】：武汉大学计算机学院;
【基金】：国家自然科学基金(61472291,61303115)
【分类号】：TP391.3

【相似文献】