Telecom Fraud Based On Support Vector Machine Prevention Analysis And Research

With the rapid development of China’s economy and telecommunications industry,the illegal and criminal acts of using telecommunications technology to commit fraud crimes spread rapidly like a virus.In recent years,non-contact fraud crime has become the main means of fraud crime,primarily through the communication tools such as mobile phones,fixed phones,and the Internet etc.causing huge losses to the people.However,due to the existence of vacuum areas in industry supervision,high investigation costs,and complicated inquiry procedures,the current fight against Telecom fraud is in trouble.How to analyze and study Telecom user numbers through intelligent means,effectively identify Telecom fraud users and strangle Telecom fraud in the cradle is the focus of current research.Based on the above problems,this paper studies the application of support vector machine in Telecom fraud prediction.Firstly,the original bill data in the telecom data is selected as the call record bill record data of telecom users,supplemented by the user’s IT information to screen the prediction indicators of Telecom fraud.Secondly,in terms of technology,this paper takes the prediction accuracy and average fitting deviation of the model as the standard to measure the prediction performance of the model,constructs an appropriate feature vector through feature engineering,and reduces the dimension of the input variables with the help of principal component analysis.To eliminate the multiple collinearity variables and improve the speed of the model,support vector classification machine and support vector regression machine are constructed to predict Telecom fraud combined with regression fitting,and the parameter optimization method of particle swarm optimization algorithm is tried to optimize the prediction performance of the model as much as possible.Finally,the multi-core support vector machine is constructed to further improve the effect of the model,so as to get a better prediction effect of Telecom fraud.Through analysis and research,it is found that the dimensionality reduction through principal component analysis can greatly improve the running speed of support vector machine model with slight loss of prediction accuracy;In terms of parameter optimization,particle swarm optimization algorithm is better than ordinary grid search method in three directions:model iteration times,running time and prediction accuracy;On the other hand,multi-core support vector machine can improve the performance of Telecom fraud prediction model at the expense of acceptable speed.Through the research of this paper,we hope that the continuously optimized support vector machine model can bring some guidance and reference to telecom enterprises and government departments in predicting Telecom fraud.