Study On The Identification Of False Trading In Big Data Environment

With the development of e-commerce platforms,more and more consumers take comments as an important reference when they choose products,and then there is the phenomenon of click farming for profit.Therefore,how to identify false trading efficiently and accurately has become a major research topic.In this paper,starting from the transaction data,modeling is divided into two aspects,and the feature representation and classification based on the depth neural network model are studied in order to solve the problem of false trading identification.In terms of text content features,an algorithm based on Bi-LSTM + Attention is used for classification.After preprocessing steps such as cleaning,word segmentation and word embedding,the data is input into Bi-LSTM to extract the feature vector,the weight is obtained by attention mechanism and the feature vector is updated.Finally,the data is classified by the fully connected layer and the softmax function.Bi-LSTM can solve the long-distance dependence of text data and consider the context information more comprehensively.The attention mechanism can optimize the text feature vector by weighting the key features.In terms of behavior features,an algorithm based on CNN is used for classification.Discrete data is optimized for feature design through preprocessing and Embedding,because embedding can improve the accuracy of the model.It is input to CNN for convolution,pooling and other operations to extract features.Finally,it is classified by a classifier.CNN can combine features better in local parallel feature extraction,automatically and efficiently,and avoid the subjectivity and limitations of manual feature extraction.Combined with the features of consumer behavior and text content,the false trading identification model in big data environment is designed and implemented.After extracting the features,two parts of feature vectors are obtained.The two parts are spliced to obtain a joint feature vector containing both text and behavior features.Compare different classifiers to choose the best.Compared with the recognition algorithm for review text alone and other recognition methods such as traditiona machine learning,the algorithm proposed in this paper uses features more comprehensively and improves the accuracy of false trading identification.The algorithm is implemented by python.The authoritative data set published on the Internet is used to independently verify the classification model.Crawler technology is used to crawl real transaction data from Taobao platform for false trading identification and verification.The neural network model is trained by adjusting the parameters,and different classification algorithms are tried to select the optimal classification.The results show that the accuracy and recall rate of this algorithm are better than those similar algorithms.This paper has the research results of false trading identification,which proves the rationality of the model and the validity of the corresponding algorithm,and provides an effective research and identification method for the study of false trading identification.