基于TensorFLow的个性化推荐系统设计

Design of Personalized Recommendation System Based on TensorFLow

文章基于TensorFLow设计了个性化推荐系统,系统循环神经网络模块可就时间序列构建模型,充分挖掘用户不断变化的兴趣爱好,而系统训练模块可就TensorFLow结构通过数据流图构建模型,基于Spark集群并行训练模型,从而调节多超参数。通过系统实现证明,此系统可实现多超参数调节,在很大程度上节省训练时间,且能显著降低误差率,动态化效果良好,满足了用户的多元化与个性化需求,值得大力推广与广泛应用。

In this paper,a personalized recommendation system is designed based on TensorFLow.The system cyclic neural network module can construct a model for time series and fully mine the changing interests of users,while the system training module can construct a model through data flow graph for TensorFLow structure and a parallel training model based on Spark cluster,so as to adjust multiple superparameters.Through the implementation of the system,it is proved that the system can realize multi-superparameter adjustment,greatly save training time,and can significantly reduce the error rate,and the dynamic effect is good,which meets the diversified and personalized needs of users,and is worth popularizing and widely used.