Comparison with Parametric Optimization in Credit Card Fraud Detection

Survey on Fraud Detection Techniques Using Data Mining

“…Our experiments are consistent with the premature result of Maes, which conclude that BN is better than NN, by comparison of result. Cost sensitive training significantly improves the performance of all categorization methods separately from NB and, separately of the DT, training mode and AIS with, optimized parameters, is the best methods in our experiments [10].…”

“…We apply five classification methods, Bayesian Nets (BN), Neural Nets (NN), Naive Bayes (NB), Artificial Immune Systems (AIS) and Decision Trees (DT), to detection the credit card fraud [10]. For a fair relationship, we fine adjust the parameters for each technique either through full search, or through Genetic Algorithm (GA).…”

Data Mining with Skewed Data

“…Cost-sensitive procedure Elkan (2001) replicates (oversampling) the minority (fraud) class according to its cost in order to balance different costs for false positives and false negatives. In Gadi et al (2008a) we achieved interesting results by applying a cost-sensitive procedure. Two advantages of a good implementation of a cost-sensitive procedure are: first, it can enable changes in cut-off to the optimal cut-off, For example, in fraud detection, if the cost tells one, a cost-sensitive procedure will consider a transaction with as little as 8% of probability of fraud as a potential fraud to be investigated; second, if the cost-sensitive procedure considers cost per transaction, such an algorithm may be able to optimise decisions by considering the product [probability of event] x [value at risk], and decide on investigating those transactions in which this product is bigger.…”

Hybrid approaches for detecting credit card fraud

“…The rule-based approach is successful for identifying fraudulent transactions that follow previously observed fraud patterns, but it lacks agility. Before a new rule is added to the existing rule set, a considerable number of fraudulent transactions matching that rule have typically already Gadi, Wang, & Lago, 2008aGadi, Wang, & Lago, 2008bPatil, Karad, Wadhai, Gokhale, & Halgaonkar, 2010Sherly & Nedunchezhian, 2010Bhattacharyya et al, 2011Sahin & Duman, 2011aAlowais & Soon, 2012 Neural networks (NN) Maes, Tuyls, Vanschoenwinkel, & Manderick, 1993Aleskerov, Freisleben, & Rao, 1997Gadi et al, 2008aGadi et al, 2008bSahin & Duman, 2011b Bayesian networks (BN) Maes et al, 1993Filippov, Mukhanov, & Shchukin, 2008Gadi et al, 2008aGadi et al, 2008b Naïve Bayes (NB) Filippov et al, 2008Gadi et al, 2008aGadi et al, 2008bAlowais & Soon, 2012 Support vector machines (SVM) Chen, Chen, Chien, & Yang, 2005Bhattacharyya et al, 2011Sahin & Duman, 2011aHejazi & Singh, 2012 Genetic algorithm (GA) Ma & Li, 2009Ozcelik, Isik, Duman, & Cevik, 2010Duman & Ozcelik, 2011 Artificial immune system (AIS) Gadi et al, 2008aGadi et al, 2008b Hidden Markov model (HMM) Bhusari & Patil, 2011Rani, Kumar, Mohan, & Shankar, 2011 occurred. A long delay is required before a rule can be added, during which time fraud strategies may change, making the rule obsolete (Krivko, 2010).…”