基于可变精度的协同过滤推荐算法研究

基于可变精度方法设计了新的协同过滤算法。首先,选取了电影评价数据集,利用可变精度方法得到新的数据集合,经过协同推荐给出了推荐结果;其次,通过平均绝对误差(MAE)和F-mean两种指标对推荐结果精度进行了评价,发现随着可变精度的提高,推荐性能得到提升,但精度需要控制在一定阈值范围内,否则超过阈值推荐性能反而下降。最后,还利用社交网络上糖尿病微博数据收集对此方法进行了实证研究,从而验证了所提方法的有效性和良好扩展性。

Collaborative Filtering Recommendation Algorithm Based on Variable Precision

In this paper, we design a new collaborative filtering algorithm based on variable precision method. Firstly, we choose movie rating data as data bed and gain new data sets using variable precision. Collaborative filtering engine is used to give recommendation result. Next, we take mean absolute error (MAE) and F-mean as metrics to measure the recommendation results. We find recommend performance better and better with the increase of variable precision. But the precision should be controlled in a certain threshold. If precision exceeds the threshold, on the contrary, the recommendation performance decline. Finally, using this method, we empirical study social media data collection on diabetic. Thereby, we verify the effectiveness and good scalability of this method proposed.