SmartScore®
Adaptive Behavioural Profiling to enhance your fraud strategy performance.
- Enhanced fraud detection
- Reduced false positive rates
- Support for your existing models and tooling
SmartScore® complements the aiRiskNet® solution by providing behavioural risk scores for use in aiRiskNet® fraud rules, utilising the latest machine learning technology.
The machine learning model scoring increases fraud detection performance while reducing false positives. Simple scoring APIs enable the rapid development and deployment of new fraud detection techniques without requiring software updates, leading to massively reduced development effort and fraud detection improvements.
Key Benefits
-
Automated model generation and implementation using aiAutoPilot ML®
aiAutoPilot ML® automatically trains machine learning models and selects the best options for your current fraud strategy. The model can then be automatically exported into SmartScore® once signed off by a fraud manager. -
Specialised in-house models are supported out of the box
SmartScore® supports a wide range of machine learning model formats, enabling your data scientists to continue using their everyday tools for model development. -
API-based flexibility and extensibility
Fraud strategy managers often supplement rules with external, manual, data-driven processes. The API extensions enable these manual processes to be implemented into a simple API, which can then be evaluated with live data. The API enables aiRiskNet® to quickly integrate with third-party services such as dark web scanning, third-party fraud models, and emergency fraud trend detection mechanisms. -
Save time
The modelling process enables the resulting model to learn fraud patterns in the data, saving time usually spent developing manual rules and resulting in better performance. -
Less complexity
A machine learning model can often replace thousands of manual rules. This means reduced false positives, increased fraud capture and reduced fraud strategy management manual effort. -
Discover SmartScore®
Book your demo today.
Features of SmartScore®
Model agnostic scoring
SmartScore® supports models built in a wide variety of machine learning frameworks, including support for custom models, providing complete flexibility for a variety of scoring needs. Have your own model you want to use? Integration is easy.
AutoML Supported
AutoML solutions such as Azure ML allow simple model development and integration with SmartScore®.
AI powered decision engine
As well as scoring fraud detection models, SmartScore® is also able to utilise non-fraud models to enable rapid business insight and decisioning. This uses the online payments data in real time.
How it works
Machine learning models trained via aiAutoPilot ML® or other tools, such as cloud-based AutoML solutions, are supported.
Models can either be imported directly or utilised using AutoML’s scoring APIs instead. Once the models are imported into SmartScore®, they can then be used in fraud rules for immediate use.
Flexible options to fit your needs
-
Build the model yourself to have complete control
SmartScore® can ingest other models. Your data scientists can use tools such as Azure ML, Tensorflow, Scikit-learn, or Pytorch to build behavioural models in-house and deploy them into SmartScore®. There is no need to adapt to new ML processes to accommodate loading models into SmartScore® - simply export your model directly into SmartScore®. -
We’ll build the model for you
Leverage ai’s skills and experience in fraud detection, and have our team build a behavioural scoring model tailored to your business to optimise fraud detection performance. -
Use aiAutoPilot ML® to automate the model-building process
Benefit from our proprietary solution, aiAutoPilot ML®, a machine learning fraud strategy optimisation tool that automates the entire fraud management process. This option will save you time and precious resources while building optimal fraud detection models for your fraud strategy.