How approval decision made in split second can increase the

Solution

Nowdays, the bank transactions are secured using a model or a desicion making system that helps to avoid losses due to intervention of fraud community in bank transactions.

The advanced non-linear statistical models were introduced so as to detect fraud in real time environment. Nonlinear models like neural networks have the abiltiy to understand and include interactions between various types of behaviors. Most risk phenomena are non-linear in their relationship to the target and most important being Fraud that is quickly changing as fraudsters are getting more and more sophisticated and talented.

The use of neural network-based behavior models in real time has literally changed the face of fraud mangement all over the world. It significantly reduced bank\'s fraud exposure in areas where there is a need to react in split seconds and stop the transaction before money goes into the hands of the fraudster.

When we examine credit card/ debit card based transactions happening at a merchant point of sale (POS) terminal, then we realize that the bank has only few milliseconds to approve or decline the decision on the transactions.

Example:

Consider a customer is shopping from online e-commerce shop and shopping done is for 2000 $. The bank has just seconds to whether approve or decline the transaction. Considering the amount of time it takes to send the decision back to the host system and so on, the time available to decide for fraud transactions is order of milliseconds. So the model must be very fast at the production system so as to judge or decide the transaction\'s nature. Hence the production execution of the decision model has to precise (from a fraud detection perspective) and extremely fast in terms of providing answers. The data driven fraud detection has transformed the sense of security in the banking industry.