Research On Mobile Phone False Numbers And Telecomunication Fraud Detection Technology Based On Operator's Data

With the rapid development of mobile communication technologies and applications,social criminals use mobile phone,SMS and other means of communication for telecommunication fraud,which has become a major problem that jeopardizes user security.Furthermore,false selling number is a telecommunication agent for profit purposes,activating the number privately and disguise the number as being in use which is a waste of marketing costs,a reduction in work efficiency,and an increase in management difficulty.Therefore,the research on the detection technology of false selling number and telecommunication fraud is of great significance.The research focus of this paper is on analyzing and modeling telecommunication data,and using this information to detect false selling number and telecommunication fraud.The main contributions of this paper are as follows:(1)A false selling number detection model is proposed.The model is mainly composed of two parts:a statistical learning module and a machine learning module.First,through the statistical learning module,a certain number of positive samples of the false selling number users are obtained,and the positive samples are used to assist the machine learning module for training.Based on the K-means algorithm,the machine learning module finds users who are suspected of false selling number by clustering,and improves the recall rate of detection.After verification based on the actual telecommunication dataset,the model can effectively detect users with false selling number.(2)A telecommunication fraud user detection model is proposed.The model first preprocesses the telecommunication dataset,which contains the common user number and the telecommunication fraud user number provided by the Public Security Bureau's Fraud Telephone Reporting Center or Tencent Security Manager.Then based on GBDT,random forest,Adaboost,XGBoost and other classification algorithms for training and prediction,a telecommunication fraud detection algorithm based on multi-classifier fusion algorithm is proposed,which can effectively improve the detection performance.Finally,the algorithm models are evaluated.The evaluation results show that the fusion algorithm has advantages in accuracy compared with the traditional classification algorithm.(3)A 1D-CNN combine algorithm combining the one-dimensional convolutional neural network with the multi-classifier fusion classification algorithm is proposed.After verification of telecommunication data,the algorithm has better prediction results than traditional algorithms,and has obvious advantages in evaluation criteria such as accuracy,recall rate and average accuracy.Finally,it is introduced how to apply the model to the actual work of telecommunication operators,and prove the validity and feasibility of the proposed model in solving practical problems.Based on telecommunication data,the paper studies data mining technology,designs and implements false selling number detection model and fraud number detection model.The research results of this paper provide theoretical support for the prevention and control of false selling of phone number and telecommunication fraud in the telecommunication industry,and also provide some ideas for the application of data mining in other industries.