个性化推荐系统中相似度计算的优化

Optimized similarity calculation in personalized recommendation system

个性化推荐系统能够比较有效的解决我们获取信息时遇到的信息过载问题,发展至今产生了许多经典的推荐算法,其中最成熟应用最为广泛的是协同过滤算法。相似度的准确计算在协同过滤算法中起到了非常重要的作用,为了进一步提高推荐系统的推荐准确率,本文对相似度计算方法进行了研究。通过项目相似度和评分差异性对计算结果影响的大小,计算时给予不同的权重,并在Movie Lens上对推荐结果进行预测,试验结果显示,MAE值降低了2.5%,优化后的相似度计算方法可以提高个性化推荐系统的推荐准确率。

Personalized recommendation system can effectively solve the information overload problem.Many classic recommendation algorithms have came into our world with the development of the recommendation system technology. Collaborative filtering algorithm is the one most widely used.Similarity plays a very important role in collaborative filtering algorithm. In order to improve the recommendation accuracy of recommendation system,we research the similarity calculation method in this paper. We optimize the traditional similarity calculation method,considering item similarity and difference in score,give different weights when we calculate the similarity. At last,we forecast the result of the recommendation on the data set named Movie Lens. The result of the exprement shows that the value of MAE decreases by 2.5% and the optimized method increases recommendation accuracy.