一种RNN-DBN的网络购物风险评估方法  

A Method of Online Shopping Risk Assessment Based on RNN-DBN

作　　者：曲媛媛 宫莉莹 贺维 QU Yuan-yuan;GONG Li-ying;HE Wei(School of Computer Science and Technology,Harbin University of Science and Technology,Harbin 150080,China;School of Information,Heilongjiang Agricultural Engineering Vocational College,Harbin 150088,China;School of Computer Science and Information Engineering,Harbin Normal University,Harbin 150025,China)

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,黑龙江哈尔滨150080 [2]黑龙江农业工程职业学院信息学院,黑龙江哈尔滨150088 [3]哈尔滨师范大学计算机科学与信息工程学院,黑龙江哈尔滨150025

出　　处：《哈尔滨理工大学学报》2019年第4期105-109,共5页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(61773388,61374138)

摘　　要：针对网络购物过程中的交易风险问题,提出一种利用深度学习技术中的循环神经网络(recurrent neural network,RNN)模型和深度置信网络(deep belief network,DBN)模型来进行网络购物风险评估的方法。该方法首先确定交易风险评估的多个影响因素,然后采用RNN模型对主观因素进行语义分析和情感分类,从而实现定性的主观评价到定量的客观评价的转化,最后采用DBN模型对所有客观影响因素进行交易风险综合评估。通过模拟实验验证,所提出的方法能够有效的解决交易风险评估问题,同时相比传统方法准确性更高,且评价结果更为科学。Aiming at the problem of transaction risk in online shopping,this paper proposes a method of online shopping risk assessment based on recurrent neural network(RNN)model and deep belief network(DBN)model in deep learning technology.Firstly,the method determines multiple influencing factors of transaction risk assessment.Then we use the RNN model to carry out semantic analysis and sentiment classification of subjective factors so as to realize the transformation from qualitative subjective evaluation to quantitative objective evaluation.Finally,the DBN model is used to analyze all the objective influencing factors conduct a comprehensive assessment of transaction risk.The simulation results show that the proposed method can effectively solve the problem of transaction risk assessment,and at the same time,it has higher accuracy compared with the traditional method,and the evaluation result is more scientific.

关 键 词：深度学习 循环神经网络 深度信念网络 风险评估 网络购物 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]