Font Size: a A A

Research And Implementation Of "Spammer" Discrimination System Based On Deep Learning

Posted on:2022-01-31Degree:MasterType:Thesis
Country:ChinaCandidate:H M YangFull Text:PDF
GTID:2518306347955929Subject:Master of Engineering
Abstract/Summary:PDF Full Text Request
With the rise of e-commerce,online shopping has gradually replaced offline shopping activities.Due to insufficient supervision of e-commerce platforms,merchants have hired a large number of spammer to comment on products in order to obtain the greatest benefits.So that more and more false product reviews are presented to consumers.This phenomenon makes it impossible for consumers to purchase goods that meet their needs through the information evaluated by other users.Therefore,the deep learning technology is combined to realize the learning of commodity review content,and the discrimination of spammer is realized through the learning results.The main work of this paper is as follows:(1)Acquisition of data sets through web crawler technology,and carry out a series of relevant preprocessing and manual labeling of data;(2)A deep learning based on spammer identification model-BI-LSTM+CNN model is proposed.The model combines the text characteristics and characteristics of user behavior,use bidirectional long and short time memory network to extract contextual semantic information of comment content,use Convolutional Neural Networks(CNN)to extract user behavior features and combine the features with asynchronous long convolution kernels,to improve the recognition effects of the spammer;(3)Introducing attention mechanism into Bi-LSTM+CNN network model,introduce the Attention mechanism for each output node before and after the bidirectional long and short time memory network,and add the attention module to the convolutional layer of the CNN model.The attention mechanism calculates the weights of features and adjusts the weights of features that have a greater impact on the classification results,to improve the accuracy of the model.Meanwhile,the attention mechanism can visualize the weight of the output features,this improves the confidence of the spammer recognition;(4)Use the spammer identification model proposed in this paper,to design and construct the spammer identification system,and realize functions such as the identification of the spammer and the management of user information.
Keywords/Search Tags:Internet marketer identification, Bi-LSTM, Convolutional neural network, Attention mechanism
PDF Full Text Request
Related items