基于评分和项目特征的群组推荐方法

Group recommendation method based on ratings and item features

针对现有群组推荐方法准确率偏低的问题,提出了一种基于评分与项目特征相结合的方法。首先综合时间因素对评分的影响和项目领域特征,利用改进的TF-IDF方法构建成员在各个特征上的偏好模型;然后考虑群体用户间的相互作用,从项目特征属性均值相似性权重和特征属性频度权重两个方面来得到群体偏好模型;最后计算群组在项目特征和评分上的综合相似度,进行预测评分并推荐。通过在Movie Lens数据集上进行实验,表明本方法比现有方法的准确率有明显提高。

To solve low accuracy problems of existing group recommendation methods, this paper proposed a new method based on ratings and item features. Firstly, considering effect of time factor on the score and item features, it used the improved TF-IDF method to build preference model for group members in the various features. Then, it considered the interaction between users in group, two aspects of the mean of item feature similarity weights and item feature frequency weights to get group preference model. Finally, it calculated comprehensive similarity for group on item feature and the score, forecast rates and recommend items. Experimental results with MovieLens data set shows that the method has significantly improved over the accuracy of existing methods.