MIT researchers use machine learning for credit card fraud detection

Using machine learning, the system can reduce false positive predictions by 54% over traditional models


Researchers from the Massachusetts Institute of Technology (MIT) have created a new system of machine learning (ML) named Deep Feature Synthesis (DFS) which can effectively detect credit card fraud.

Current fraud-detecting systems have a habit of flagging up sales as suspicious, meaning that customer's credit cards are often declined when using them at a new store or location. After being fed data from 1.8 million transactions from a large bank the DFS system, based on "automated feature engineering", indicated a 54% reduction in false positive predictions compared to traditional models.

Visit DATAx New York on December 12–13, 2018

"The big challenge in this industry is false positives," said Kalyan Veeramachaneni, a principal research scientist on the project.

"We can say there's a direct connection between feature engineering and (reducing) false positives... That's the most impactful thing to improve accuracy of these machine-learning models," Veeramachaneni added.

The system extracts highly detailed features from data generated from around 133,000 false positives and compares them with 289,000 false positives. Using information extracted from over 200 features for each transaction means that when a user swipes a card the model checks whether the features match fraud behavior. If they do, the sale would be blocked.

When utilized, the technology will save bank's money and improve overall customer experience.

Visit Innovation Enterprise's Machine Learning Innovation Summit in Dublin on November 29, 2018

3 tips to help your ar or vr product succeedsmall

Read next:

3 tips to help Your AR or VR product succeed