基于AutoRec的推荐系统模型研究

Research on recommendation system model based on AutoRec

传统推荐算法对用户的历史记录有严重的依赖,同时面临工程复杂与表达能力差,挖掘数据潜藏模式难的问题。为增强模型表达能力,满足更多数据类型和业务场景,以及解决推荐系统冷启动问题,实验采用单隐层神经网络推荐模型AutoRec进行改进,用于计算用户User与物品Item之间隐藏特征。同时,采用概率统计学的方法,解决系统冷启动问题。实验在PyTorch框架上进行数据集的训练,与传统推荐算法相比,该模型具有更强的表达能力和更精准的推荐结果。

The traditional recommendation algorithm relies heavily on the historical records of users,and at the same time,it is difficult to mine the hidden patterns of data due to the complexity of engineering and poor expression ability.To enhance the ability of model expression,meet more data types and business scenarios,and solve the cold startup problem of the recommendation system.In this experiment,AutoRec,a recommendation model of single hidden layer neural network,was used to calculate hidden features between User and Item.At the same time,the method of probability statistics is used to solve the cold start problem of the system.In this experiment,data sets are trained on PyTorch framework.Compared with traditional recommendation algorithms,this model has stronger expression ability and more accurate recommendation results.