AI, Supervised and Unsupervised Machine Learning

AI, Supervised and Unsupervised Machine Learning

Making transaction data actionable with modern AI algorithms and supervised and unsupervised machine learning models

How does INETCO use AI, supervised and unsupervised machine learning for pattern recognition and scoring?

The INETCO Insight machine learning engine is uniquely optimized for real-time payment data. Some of these optimizations include:

  • A real-time data pipeline
  • High speed transaction data store
  • Real-time transaction tracing and end-to-end correlation
  • Rules-based alerts engine
  • On-demand analytics engine
  • In-memory execution framework

INETCO Insight matches the right AI and machine learning algorithms to the problems and requirements outlined, such as cash forecasting, predicting card usage or customer buying habits, and scoring transaction risk in milliseconds. Self-learning machine learning algorithms are easily configurable and built on newer algorithms such as Isolation Forest and Gradient Boosting, incorporating seasonality, individual real-time transaction events and in-depth transaction data to increase precision. Both supervised and unsupervised machine learning algorithms are used in conjunction to assess the validity of a transaction.

 

How INETCO Insight machine learning capabilities work

Continuous real-time transaction data feed for machine learning models

Independently collect real-time transaction data straight off the network. This in-depth transaction data set is complete with all message fields – nothing is stripped off. More complete data means more advanced AI and machine learning models – pertinent intelligence can be extracted and continuously fed into these models to increase the speed and precision of detection.

Supervised Machine Learning Models

Automatically learn from labeled fraud cases and detect fraudulent behavior patterns based on previously confirmed fraud cases. Decrease the amount of transaction anomalies mistakenly detected and reduce customer frustrations due to false positives. Increase the efficiency of fraud analysts and make their job a lot easier.

Unsupervised Machine Learning Models

Build individual customer models on the fly and flag anomalies per customer, card or device based on past behavior. Detect new fraudulent patterns that have not previously been seen. Different weights can be assigned to the extracted data points to help you fine tune the model so that an optimal balance between risk, cost and convenience is attained.

“The additional visibility INETCO Insight has provided has allowed Solutran to understand where blockages can occur and what is happening at a specific endpoint during a transaction rejection.”

JASON PRIGGE, CTO AT SOLUTRAN

Resources