论文部分内容阅读
针对当前主流推荐算法无法甄别离群样本和弱贡献率样本,且单模型算法泛化能力较弱等问题,提出一种基于级联过滤的多模型融合的推荐方法.该方法先采用级联回归模型过滤掉离群样本和弱贡献率样本;然后,把推荐问题抽象成二分类问题和回归问题,分别采用基于Bagging的随机森林和基于Boosting的梯度提升回归树两种树型算法、线性的逻辑回归算法来拟合用户兴趣;最后,将这三种算法分别训练若干模型进行线性融合,取Top-N推荐.实验表明,该方法不仅有效提高了推荐精度,还增强了模型的泛化能力,具有较强的实用价值.
Aiming at the problems that the current mainstream recommendation algorithm can not identify outlier samples and weak contribution rate samples, and the single model algorithm has weak generalization ability, a multi-model fusion method based on cascade filtering is proposed. This method uses cascade regression The model filters the outlier samples and the samples of weak contribution rate. Then, the proposed problem is abstracted into the dichotomous problem and the regression problem. Two tree-based algorithms, Bagging-based random forest and Boosting-based gradient enhancement regression tree, Regression algorithm to fit the user’s interest.Finally, the three algorithms are respectively trained on a number of models for linear fusion, taking Top-N recommendation.Experiments show that this method not only effectively improves the recommendation accuracy, but also enhances the generalization ability of the model, Has strong practical value.