Operationalization—or putting models to work—is an important part of the data science lifecycle. You might have trained and validated the most accurate model in the world, but if it just stays in the development folder of RapidMiner Studio, it’s not going to deliver much value, is it? Predictive models deliver value when we put them to work and make accurate predictions to mitigate risks, drive customer satisfaction, and help businesses grow.
Sometimes operationalization means periodically applying the model to large amounts of data records to score the risk associated with a portfolio, predict the probability of a user responding to a marketing campaign, or whatever other use case. This can be done by scheduling scoring processes in RapidMiner Server (now RapidMiner AI Hub).
Other times predictions need to be interactive, where users can exchange information back and forth between RapidMiner and another application like a data visualization tool. You can do this today by creating a web service for the model in RapidMiner AI Hub.
Lastly, there’s an important use case when predictions need to be delivered in real time. A personal recommendation in a web site—a delivery prediction in a phone app—an immediate risk analysis needed by a bank application. In these cases, there’s a need for very low latency and high throughput. That’s what we’ve built with the new RapidMiner Real-Time Scoring.
Key Features of
RapidMiner Real-Time Scoring
The RapidMiner Real-Time Scoring Agent extends RapidMiner AI Hub with a lightweight execution engine designed for specific use cases where speed and volume are critical. Here a few things we’d like to highlight:
- Real-time scoring agents can indefinitely scale out for demanding use cases, and are easy to deploy and easy to manage.
- The request latency for retrieving a score is less than 25ms.
- It uses a modern REST API with JSON as request and response format. It can be called from any application, anywhere.
- While scoring is the primary use case, any arbitrary process can be deployed in the Scoring Agents. If you need some preparation of the data, filtering, post-processing or anything else, it can be done.
With the new RapidMiner Real-Time Scoring, you can turn your predictions into prescriptions, even for the most demanding use cases.
If you’d like to learn more about what RapidMiner can do for your business, request a demo today.
