摘要
Abstract
DDM model introduced into the currently popular LDA model the residual component to make up for the loss of the discrimina-ted information. Because the vocabulary of document obeys the power-law distribution, residual component will be affected by the high frequency theme vocabulary. This paper puts forward HDDM model, using the improved TF-IDF function, sets the sampling weight of the residual component, and improves the low frequency vocabulary’s function in residual component. Experiments show that this model improves the quality of personalized recommendation, further promoting the precision.关键词
个性化推荐/DDM模型/HDDM模型/TF-IDF函数Key words
personalized recommendation/DDM model/HDDM model/TF-IDF function分类
信息技术与安全科学
