基于修正相似度的微博社会化推荐系统构建方法研究

Construction Method of Micro-blog Social Recommendation System Based on Modified Similarity

将影响社会化推荐的三种因素分别量化,建立了与微博社会网络一一映射。然后,基于Karhunen-Loéve(KL)变换方法,计算出了同一主题下积极性和消极性文本平均距离。最后,将社会网络信息与情感相似结合形成修正的情感相似度量方法,利用修正相似度方法构建了新的社会化推荐系统。基于微博数据的实证计算和分析显示:经过变换后的用户相似度可以得到不同程度的提高;利用修正相似度方法构建的微博社会化推荐系统更符合用户心理偏好。

Three factors that influence social recommendation system are defined into quantity and micro-blog social network mapped. Further,Positive and negative text average distances are calculated based on Karhunen-Loéve（ KL） transformation method. Finally,the new framework of social recommendation system is built. The experimental verification results using micro-blog data reflect that user similarities are increased with different degrees with KL transformation and the social recommender system based on sentiment similarity is better to catch users＇ preferences.