基于Adaboost算法的推荐系统评分预测框架

Adaboost-Based Framework For Rating Prediction in Recommender System

在机器学习领域,Adaboost算法的实用性和有效性早已被证明.然而该算法原本是为分类问题设计,因而在推荐系统领域研究问题中无法直接应用,对其应用研究相对较少.本文对Adaboost算法进行改进,通过引入阈值,将评分预测问题转化为分类问题,并利用其权重更新的思想训练模型,提出了一个针对评分预测问题的框架,可以将训练出的多个模型集成起来得到最终的评分预测,提高了预测精度.我们选取矩阵分解模型作为基本模型,实验结果表明,使用该框架可以有效提高预测精度.

In the field of machine learning, the practicality and effectiveness of the Adaboost algorithm has already been demonstrated. However, since this algorithm is originally designed for classification problems, it cannot be applied directly to rating prediction problems in recommender system field. Thus the research in this area is limited. In this paper, we improve the Adaboost algorithm. By introducing the threshold value, we transform rating prediction into classification. By updating weights in the training process, we propose a framework for the rating prediction, which can integrate the multiple training models. The final rating is obtained through the integrated model. We select the Matrix Factorization model as an instance, and the experimental results show that the framework can effectively improve the prediction accuracy.