FRAUD ACCOUNTS CLASSIFICATION MODELLING ON MULTI E-COMMERCE PLATFORM TO PREVENT CYBERCRIME

• Main Author: Sugiharto, Grawas
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/54517
• Institution: Institut Teknologi Bandung

Nowadays, cybercrime is increasingly prevalent in society. Based on data compiled by the Indonesia National Police, the number of cybercrime increases by 6.46% annually, with online fraud as the most reported crime with 7.892 cases or 44.40% out of the total cases handled. Losses due to this crime in the year 2019 reached Rp Rp 235,937,867,634.50, which occurred on four platforms, namely email (1.92%), website (13.09%), telecommunication (28.66%), and social media (56.33%) with the fraudulent modus operandi of selling goods at much lower prices below the market price. Fraud crime in e-commerce has evolved into organized crime, where the perpetrators manipulate data in such a way as to gain the trust of the victims. Therefore, it is necessary to have a common detection model for fraud perpetrators' accounts on various e-commerce platforms so that people can avoid online fraud. Modeling is performed using the Naïve Bayes classification algorithm, Decision Tree, and K-NN with different data ratio variables. From the modelling test result, the green platfrom achieved the best performa using KNN algorithm with the highest accuracy score is 90.51%; the red platform achieved the best performa using Decision Tree algorithm with the highest accuracy score is 96.89%; and multi platform achieved the best performa using Naïve Bayes with the highest accuracy score 90.02%