Application of adaboost algorithm and immune algorithm in telecommunication fraud detection

Abstract

Fraud detection is one of the biggest challenges facing the telecommunication industry now and, in the future, the fight against fraud and anti-fraud has also reached a new stage. Finding a time-sensitive fraud detection method is an important way for operators to solve this problem. In this article, to solve the low accuracy of the general algorithm, an adaptive improvement algorithm is proposed. The common algorithms are combined and enhanced, which greatly improves the accuracy of the detection results. Here we take the artificial immune algorithm as an example. The main application is combining machine learning and immune algorithm to apply to telecommunication fraud detection. The combination of the two is more conducive to pushing research on telecommunication fraud to a new stage for the future telecommunication industry.

Research on Evidence Issues in Telecommunications Network Fraud Cases

“…In the case of less evidence, the amount of fraud and personal income of the accomplice should be confirmed by the above-mentioned corroboration methods and rules of thumb, and the size of the role of the accomplice and the degree of subjective malice should be judged. The process of confirmation should be combined with the standard of sufficient evidence, the principle of favoring the defendant, and the judicial rules such as the benefit of the doubt, to make an accurate determination of the scope of the defendant's culpability [12]. In this way, the circumstances of the crime of the accessory can be more accurately determined, and the sentencing is more scientific and reasonable.…”

Legal system-oriented telecom fraud detection, identification and prevention

“…Kashir et al [11] used call detail records of normal and fraudulent users as input and used neural networks and support vector machines (SVM) for classification and found that Bayesian regularization had the best performance. Wu et al [12] proposed a method that combines machine learning and area algorithms to achieve telecom fraud detection, significantly enhancing the accuracy of detection results through the integration of common algorithms. Chang et al [13] designed a centrality-oriented deep random walk method to identify key fraudulent roles in telecom fraud networks and found that this approach outperformed other baseline methods.…”