基于影响力控制的热传导算法

因特网上信息严重过载,使得用户不容易从纷繁的信息中找到适合自己的内容。如何准确地向用户推荐他们想要的信息成为急待解决的问题。热传导算法(HC)被广泛地应用于个性化推荐领域,但是它的热量传播机制不利于经历丰富的用户喜欢的流行物品得到更多的热量。因此,本文提出了基于影响力控制的热传导算法(THC)。THC引入两个参数控制度数大的用户喜欢的度数大的物品对目标用户推荐的影响。另外,本文提出利用用户对景点的各项评分及评论的情感极性来判断用户是否喜欢一个景点,还提出了一个新的指标buir以度量度数大的用户喜欢的度数大的物品出现在推荐列表中的比例。实验结果表明:适度增大的度数大的用户喜欢的度数大的物品的影响,有助于推荐出目标用户喜欢的物品,从而有助于提升推荐效果。
The overload of information on the Internet can lead to users feeling hopeless about finding the information they are seeking. Making accurate recommendations to users about the information they truly need is an urgent problem that must be addressed. The heat conduction (HC) algorithm has recently been applied in personalized recommendation technology, but its mechanism weakens the heat generated from the larger-degree itemsliked by the larger-degree users. To solve this problem, we propose an improved HC algorithm that is based on user influence control (THC). THC introduces two tunable parameters to better control the influence of larger-degree users’ preferences for larger-degree items on target users. We also consider a user’s comment scores and the sentiment polarity of a comment in a given scenario to accurately judge whether the user truly likes the given scenario. We also propose a new index, called a buir, which measures the ratio of the larger-degree items that are liked by larger-degree users on the recommendation list. Experimental results show that appropriately promoting the influence of larger-degree items that are liked by larger-degree users helps in making recommendations to target users regarding items in which they are truly interested, thereby improving the performance of the recommendation