...detection of fraud by at least 78% in our reference dataset. We also discuss how a combination of these strategies can be embedded in a Stream-based application to detect fraud in real-time. From this perspective, our experiments lead to an average processing time of 111,702ms per transaction, while strategies to further improve the performance are discussed. Keywords: Fraud Detection, Stream Computing, Real-Time Analysis, Fraud, Data Mining, Retail Banking Industry, Data Preprocessing, Data Classification, Behavior-based Models, Supervised Analysis, Semi-supervised Analysis Sammanfattning Privatbankerna har drabbats hårt av bedrägerier de senaste åren. Bedragare har lyckats kringgå forskning och tillgängliga system och lura bankerna och deras kunder. Därför vill vi införa en ny, polyvalent...
Words: 56858 - Pages: 228
...detection of fraud by at least 78% in our reference dataset. We also discuss how a combination of these strategies can be embedded in a Stream-based application to detect fraud in real-time. From this perspective, our experiments lead to an average processing time of 111,702ms per transaction, while strategies to further improve the performance are discussed. Keywords: Fraud Detection, Stream Computing, Real-Time Analysis, Fraud, Data Mining, Retail Banking Industry, Data Preprocessing, Data Classification, Behavior-based Models, Supervised Analysis, Semi-supervised Analysis Sammanfattning Privatbankerna har drabbats hårt av bedrägerier de senaste åren. Bedragare har lyckats kringgå forskning och tillgängliga system och lura bankerna och deras kunder. Därför vill vi införa en ny, polyvalent...
Words: 56858 - Pages: 228