Introduction to Pattern Recognition
Pattern recognition is the process of recognizing patterns by using a machine learning algorithm. It can be defined as the classification of data based on knowledge already gained or on statistical information extracted from patterns and/or their representation.
This process can be used to define dispute filer patterns and answer many related questions, for example:
- Who is filing disputes the most?
- Are there any patterns in the reason?
- Are there any patterns among those filing disputes?
- Are more disputes being filed against the same merchant?
- Is it a new merchant?
- For the same type of products?
Increase in payment fraud
Payment fraud is any false or illegal transaction that can be related to1:
- Credit card fraud – usually when fraudsters use a stolen credit card and buy goods online. The credit card owner disputes the charge to get back the money, the merchant can lose the money and the product.
- Denial of receiving the product – fraudster orders the product, the merchant sends it and then the thief claims that he didn’t receive the product.
- Fake returns – fraudster tries to convince the merchant that he sent the ordered product back and wants a refund.
- Triangulation fraud – related to the theft of credit card data.
- Friendly fraud – when the fraudster orders some items, pays for them, and then initiates the chargeback, trying to convince you that his card was stolen.
According to the statistics presented in the 2018 AFP Payments Fraud and Control Report, 78% of surveyed organizations were impacted by payment fraud in 2017. New benchmark data from ACI Worldwide revealed a 13 percent increase in fraud attempts in the ‘buy online, pickup in-store’ (BOPIS) channel during the 2018 peak holiday season. Based on hundreds of millions of merchant transactions, including some of the world’s leading global retail brands, BOPIS also saw consumer transactions peak at 20 percent right before Christmas, as shoppers bought last-minute gifts before the holiday2.
AI being used to decrease fraud in payments
Artificial Intelligence, particularly Machine learning, is being used to identify patterns in payment data. What is new, in the context of fraud is, how fraud attacks are getting more sophisticated and how the financial industry copes with it using new methods3.
Examples of pattern recognition in card transactions
Dispute filer patterns
- Cardholder has only made a few purchases, and now they are filling a dispute.
- Check the percentage of disputes filed versus the # of purchase transactions made. If above a certain % flag any disputes as possibly fraudulent. The % can be based on average by demographics, regions, purchase patterns (types of merchants purchased from) and more.
Merchant patterns
- For purchases completed over a particular time period:
- Can a pattern be noticed that x% of those purchases have been disputed?
- Can a pattern be noticed that x% of purchases of a particular product from a particular merchant, have been disputed?
- Can a pattern be noticed that x% of purchases of a particular product, on a particular day/time, from a particular merchant, have been disputed?
Payment Service Provider (PSP) patterns
- For purchases completed over a particular time period:
- Can a pattern be noticed that x% of purchases that went through a particular PSP, during a specific time period day/time, have been disputed?
- Can a pattern be noticed that x% of purchases that went through a particular PSP, during a specific time period day/time, have been disputed with the reason, “double charged”?
This information can simplify decisions about a number of disputes as there could have been an issue with a PSP for example, during a particular time period.
Summary
Implementation of proper Machine Learning algorithms to pattern recognition can result in:
- Fraud detection in real-time
- Deeper knowledge about customer behavior
- Improvement of payment data credibility