一种基于电影评分预测的协同过滤

A Scoring Prediction Recommendation Algorithm Based on user Self-portrait

评分预测是推荐系统的重要环节,现在大部分的评分预测是利用用户的历史评分记录来推断该用户将给某个项目打多少分.该方法利用了用户历史评分记录,没有充分利用用户或者项目属性,平均绝对误差较大.针对以上问题,构造一种基于用户自画像的评分预测协同过滤推荐算法.该算法通过计算用户之间历史评分记录的相似度和用户自画像之间的相似度,然后计算出两种相似度的权重,把两种相似度乘以各自的权重进行组合.实验结果表明,构造的评分预测算法较好的减少预测评分和实际评分之间的平均绝对误差,提高了评分预测的准确性.

Scoring prediction is an important part of the recommendation system. Currently,most scoring predictions based on the user’s historical score are to infer how many points the user will score for an item.This method utilizes the user history score record,does not make full use of user or project attributes,and the average absolute error is large. To solve the above problems,a scoring prediction collaborative filtering recommendation algorithm is constructed,which is based on user self-portrait is calculated. In the algorithm,the similarity between the user’s historical score records and the user’s self-portrait is calculated,and then the weights of the two similarities is calculated,and the two similarities is multiplied by their respective weights. The experimental results show that the constructed scoring prediction algorithm can reduce the average absolute error between the predicted score and the actual score,and improve the accuracy of the score prediction.