Construção de um modelo baseado em floresta aleatória para detecção de fraude na substituição de cartões SIM nas empresas de telecomunicações: caso de estudo - TELECOM

Abstract

Changing SIM cards is an essential operation for mobile phone users, allowing the replacement of the SIM card with another without changing the mobile phone number. At the However, this practice has been a frequent target of fraud in recent years, presenting serious security challenges for telecommunications companies. This study focuses on building and training a Forest-based model Random for fraud detection when replacing SIM cards in companies telecommunications. The research investigated in detail the process of issuing the duplicate of the mobile phone number, where he emphasized the essential systems involved, SIMConnectX and SIMLogX. Analysis of these systems revealed a robust and meticulous structure to ensure the authenticity and security of SIM swaps. However, when categorizing fraud SIM exchanges, registration and documentation flaws were identified, highlighting patterns suspicious, such as the rapid migration of numbers between provinces and correlations between the exchange of SIM and resetting the PIN in the Mobile Wallet. A dataset was provided and it was necessary to refine it during the KDD process for the modeling, cleaning, transformation and preparation techniques were implemented data. This resulted in an optimized dataset for model building. of Random Forest. Model training focused on identifying such suspicious patterns and predicting fraudulent activities. After careful adjustments to the parameters, a model was obtained effective, exhibiting an accuracy of 94% and an F1-score of 89% in fraud detection. These results offer valuable promising approaches to understanding and fraud prevention when exchanging SIM cards, highlighting the effectiveness of the Random Forest in accurately identifying fraudulent transactions (Tradução Nossa).