摘要
针对传统数据协同过滤方法存在过滤耗时长的问题,提出基于无限深度神经网络的Web大数据协同过滤方法。通过分析神经网络连接方式与结构,在时间维度上展开,研究其"无限深"的状态;为提高计算性能,调整权值,获取网络整体性能函数,利用反向传播算法确定目标函数,再使用梯度下降法更新参数,确保网络收敛到局部最小值;利用上述无限深度神经网络分别对Web大数据的权重、梯度以及相似度进行计算,评价各个数据并进行排序,将排序靠后的大数据进行过滤。仿真结果表明,上述方法能够提高用户满意度,有效解决数据稀疏问题。
Aiming at the problem of long filtering time in traditional data collaborative filtering methods, a web big data collaborative filtering method based on infinite depth neural network is proposed.By analyzing the connection mode and structure of the neural network, the "infinite depth" state of the network was studied in the time dimension;in order to improve the computing performance, adjust the weights and obtain the overall performance function of the network, the backpropagation algorithm was used to determine the objective function, and then the gradient descent method was used to update the parameters to ensure that the network converges to the local minimum;the above infinite depth neural network was used to calculate the network performance.The weight, gradient and similarity of Web big data were calculated, each data was evaluated and sorted, and the sorted big data was filtered.Simulation results show that this method can improve user satisfaction and solve the problem of data sparsity effectively.
作者
葛涵
张志勇
张红良
GE Han;ZHANG Zhi-yong;ZHANG Hong-liang(College of Liberal Arts,Beihua University,Jilin Jilin 132013,China)
出处
《计算机仿真》
北大核心
2021年第9期400-404,共5页
Computer Simulation
基金
2020年度国家社会科学基金项目(20BXW004)。
关键词
大数据
协同过滤
兴趣度
反向传播
梯度下降算法
Big data
Collaborative filtering
Interest degree
Backpropagation
Gradient descent algorithm
