Cristián Bravo

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

In the last decade, the ease of online payment has opened up many new opportunities for e-commerce, lowering the geographical boundaries for retail. While e-commerce is still gaining popularity, it is also the playground of fraudsters who try to misuse the transparency of online purchases and the transfer of credit card records. This paper proposes APATE, a novel approach to detect fraudulent credit card transactions conducted in online stores. Our approach combines (1) intrinsic features derived from the characteristics of incoming transactions and the customer spending history using the fundamentals of RFM (Recency - Frequency - Monetary); and (2) network-based features by exploiting the network of credit card holders and merchants and deriving a time-dependent suspiciousness score for each network object. Our results show that both intrinsic and network-based features are two strongly intertwined sides of the same picture. The combination of these two types of features leads to the best performing models which reach AUC-scores higher than 0.98.

Development and application of consumer credit scoring models using profit-based classification measures

This paper presents a new approach for consumer credit scoring, by tailoring a profit-based classification performance measure to credit risk modeling. This performance measure takes into account the expected profits and losses of credit granting and thereby better aligns the model developers’ objectives with those of the lending company. It is based on the Expected Maximum Profit (EMP) measure and is used to find a trade-off between the expected losses – driven by the exposure of the loan and the loss given default – and the operational income given by the loan. Additionally, one of the major advantages of using the proposed measure is that it permits to calculate the optimal cutoff value, which is necessary for model implementation. To test the proposed approach, we use a dataset of loans granted by a government institution, and benchmarked the accuracy and monetary gain of using EMP, accuracy, and the area under the ROC curve as measures for selecting model parameters, and for determining the respective cutoff values. The results show that our proposed profit-based classification measure outperforms the alternative approaches in terms of both accuracy and monetary value in the test set, and that it facilitates model deployment.

Cost-based feature selection for Support Vector Machines: An application in credit scoring

In this work we propose two formulations based on Support Vector Machines for simultaneous classification and feature selection that explicitly incorporate attribute acquisition costs. This is a challenging task for two main reasons: the estimation of the acquisition costs is not straightforward and may depend on multivariate factors, and the inter-dependence between variables must be taken into account for the modelling process since companies usually acquire groups of related variables rather than acquiring them individually. Mixed-integer linear programming models are proposed for constructing classifiers that constrain acquisition costs while classifying adequately. Experimental results using credit scoring datasets demonstrate the effectiveness of our methods in terms of predictive performance at a low cost compared to well-known feature selection approaches.

Granting and managing loans for micro-entrepreneurs: New developments and practical experiences

We present a methodology to grant and follow-up credits for micro-entrepreneurs. This segment of grantees is very relevant for many economies, especially in developing countries, but shows a behavior different to that of classical consumers where established credit scoring systems exist. Parts of our methodology follow a proven procedure we have applied successfully in several credit scoring projects. Other parts, such as cut-off point construction and model follow-up, however, had to be developed and constitute original contributions of the present paper. The results from two credit scoring projects we developed in Chile, one for a private bank and one for a governmental credit granting institution, provide interesting insights into micro-entrepreneurs’ repayment behavior which could also be interesting for the respective segment in countries with similar characteristics.

Improving credit scoring by differentiating defaulter behaviour

We present a methodology for improving credit scoring models by distinguishing two forms of rational behaviour of loan defaulters. It is common knowledge among practitioners that there are two types of defaulters, those who do not pay because of cash flow problems (‘Can’t Pay’), and those that do not pay because of lack of willingness to pay (‘Won’t Pay’). This work proposes to differentiate them using a game theory model that describes their behaviour. This separation of behaviours is represented by a set of constraints that form part of a semi-supervised constrained clustering algorithm, constructing a new target variable summarizing relevant future information. Within this approach the results of several supervised models are benchmarked, in which the models deliver the probability of belonging to one of these three new classes (good payers, ‘Can’t Pays’, and ‘Won’t Pays’). The process improves classification accuracy significantly, and delivers strong insights regarding the behaviour of defaulters.

Patterns of Piscirickettsia salmonis load in susceptible and resistant families of Salmo salar.

The pathogen Piscirickettsia salmonis produces a systemic aggressive infection that involves several organs and tissues in salmonids. In spite of the great economic losses caused by this pathogen in the Atlantic salmon (Salmo salar) industry, very little is known about the resistance mechanisms of the host to this pathogen. In this paper, for the first time, we aimed to identify the bacterial load in head kidney and muscle of Atlantic salmon exhibiting differential familiar mortality. Furthermore, in order to assess the patterns of gene expression of immune related genes in susceptible and resistant families, a set of candidate genes was evaluated using deep sequencing of the transcriptome. The results showed that the bacterial load was significantly lower in resistant fish, when compared with the susceptible individuals. Based on the candidate genes analysis, we infer that the resistant hosts triggered up-regulation of specific genes (such as for example the LysC), which may explain a decrease in the bacterial load in head kidney, while the susceptible fish presented an exacerbated innate response, which is unable to exert an effective response against the bacteria. Interestingly, we found a higher bacterial load in muscle when compared with head kidney. We argue that this is possible due to the availability of an additional source of iron in muscle. Besides, the results show that the resistant fish could not be a likely reservoir of the bacteria.

Whole-genome comparative analysis of the pathogen Piscirickettsia salmonis

The intracellular pathogen Piscirickettsia salmonis is the etiological agent of piscirickettsiosis, the most important bacterial disease that affects the Chilean salmon industry. Despite its importance, little is known regarding the biology of the pathogen. In this study, recently published sequencing data was used in order to characterize the genome of P. salmonis, defining groups of genes associated with bacterial processes such as, invasion and intracellular survival. Moreover, one Chilean P. salmonis isolate, which is known to be virulent at in vitro and in vivo assays, was sequenced, assembled, annotated and functionally characterized. Whole-genome comparisons between public P. salmonis isolates confirmed the existence of two different genogroups associated with the LF-89 and EM-90 strains, and the bacterial pan and core genome were defined. Additionally, differences were observed at the genomic level between the P. salmonis reference strain and a Norwegian isolate, which is known to produce milder piscirickettsiosis outbreaks. Finally, candidate genes for invasion and intracellular survival were chosen from phylogenetically related bacteria, and annotated in P. salmonis using comparative genomics. These results showed the presence of several genes that might be related to bacterial pathogenesis, for example those of the type III, IV and VI secretion systems, in which some amino acidic differences within both genogroups and the Norwegian isolate were established. Altogether, these results will be relevant for understanding the host-pathogen interaction and further studies, aimed at generating new disease control strategies, should be devised using this information.

A Hybrid System for Probability Estimation in Multiclass Problems Combining SVMs and Neural Networks

This paper addresses the problem of probability estimation in multiclass classification tasks combining two well known data mining techniques: support vector machines and neural networks. We present an algorithm which uses both techniques in a two-step procedure. The first step employs support vector machines within a one-vs-all reduction from multiclass to binary approach to obtain the distances between each observation and the support vectors representing the classes. The second step uses these distances as inputs for a neural network, built with an entropy cost function and softmax transfer function for the output layer where class membership is used for training. Consequently, this network estimates probabilities of class membership for new observations. A benchmark using different databases demonstrates that the proposed algorithm is highly competitive with the most recent techniques for multiclass probability estimation.

A comparative study of social network classifiers for predicting churn in the telecommunication industry

Relational learning in networked data has been shown to be effective in a number of studies. Relational learners, composed of relational classifiers and collective inference methods, enable the inference of nodes in a network given the existence and strength of links to other nodes. These methods have been adapted to predict customer churn in telecommunication companies showing that incorporating them may give more accurate predictions. In this research, the performance of a variety of relational learners is compared by applying them to a number of CDR datasets originating from the telecommunication industry, with the goal to rank them as a whole and investigate the effects of relational classifiers and collective inference methods separately. Our results show that collective inference methods do not improve the performance of relational classifiers and the best performing relational classifier is the network-only link-based classifier, which builds a logistic model using link-based measures for the nodes in the network.

Semi-supervised constrained clustering with cluster outlier filtering

Constrained clustering addresses the problem of creating minimum variance clusters with the added complexity that there is a set of constraints that must be fulfilled by the elements in the cluster. Research in this area has focused on “must-link” and “cannot-link” constraints, in which pairs of elements must be in the same or in different clusters, respectively. In this work we present a heuristic procedure to perform clustering in two classes when the restrictions affect all the elements of the two clusters in such a way that they depend on the elements present in the cluster. This problem is highly susceptible to outliers in each cluster (extreme values that create infeasible solutions), so the procedure eliminates elements with extreme values in both clusters, and achieves adequate performance measures at the same time. The experiments performed on a company database allow to discover a great deal of information, with results that are more readily interpretable when compared to classical k-means clustering.