Transaction aggregation as a strategy for credit card fraud detection

Abstract

The problem of preprocessing transaction data for supervised fraud classification is considered. It is impractical to present an entire series of transactions to a fraud detection system, partly because of the very high dimensionality of such data but also because of the heterogeneity of the transactions. Hence, a framework for transaction aggregation is considered and its effectiveness is evaluated against transaction-level detection, using a variety of classification methods and a realistic cost-based performance measure. These methods are applied in two case studies using real data. Transaction aggregation is found to be advantageous in many but not all circumstances. Also, the length of the aggregation period has a large impact upon performance. Aggregation seems particularly effective when a random forest is used for classification. Moreover, random forests were found to perform better than other classification methods, including SVMs, logistic regression and KNN. Aggregation also has the advantage of not requiring precisely labeled data and may be more robust to the effects of population drift.

APATE: A novel approach for automated credit card transaction fraud detection using network-based extensions

“…While ANN's generally achieve a high performance, they are black box models which lack interpretability. Recently, the use of ensemble methods like random forests is found to perform well in credit card fraud (Whitrow et al, 2009;Bhattacharyya et al, 2011;Dal Pozzolo et al, 2014). Random forests work especially well when there are many input features to learn from, which is often the case in network-related classification problems (Henderson et al, 2011).…”

“…There were some variables from the literature which could not be implemented in our study, given the availability of data: some works in the literature use three months of data (Bhattacharyya et al, 2011), but we only have one month available so it was impossible. We also have available only online transactions of one issuer, so bank-related and POS related variables are not applicable, such as in Sánchez et al (2009) and Whitrow et al (2009).…”

“…last week, as suggested by Bhattacharyya et al (2011) and Whitrow et al (2009), which were estimated both at global transaction level and at merchant level.…”

Survey on Credit Card Fraud Detection using Recurrent Attributes

“…Since it is not always possible or easy for a human analyst to detect fraudulent patterns in transaction datasets, constantly characterized by a large number of samples, many dimensions and online updates, automatic systems are imperative. Also, the cardholder is not reliable in reporting the theft, loss or fraudulent use of a card [5]. Since the number of fraudulent transactions is much smaller than the legitimate ones, the data distribution is unbalanced, i.e.…”

Um processo para modelagem e aplicação de técnicas computacionais para detecção de fraudes em transações eletrônicas

“…Ela também pode requerer ligações para os vendedores envolvidos, ou um trabalho ainda mais extensivo dependendo do tamanho da perda em potencial, ou seja, do valor da transação. Porém, dado que a fraude é uma ocorrência rara quando comparada à quantidade total de transações legítimas, haverá uma preponderância de falsos alarmes, sendo então importante manter o custo da investigação baixo, principalmente quando a perda em potencial for baixa (Whitrow et al, 2008). Dado o custo envolvido no processo, se o valor da transação for baixo o suciente a ponto de ser menor que o custo da revisão manual, pode até mesmo não ser viável realizar a análise da transação, mesmo que ela tenha um alto indício de fraude.…”