One of the difficulties in deploying a machine learning strategy is ensuring the performance of real-time decisions made using predictive models. The demands for reliability and speed to support real-time predictive systems have increased. Those challenges are made more difficult by the increasing size and complexity of models and algorithms used to improve the accuracy of decisions.

There are many different systems that are available to build the learning part of a machine learning pipeline, but many of these systems leave the decision making system as an exercise for the reader. Building customer services to support real-time decision making can be difficult to do reliability and with scale.

Instead of building a custom system, we will look at how Redis 4.0 and the Redis-ML module can be used out of the box to provide a real-time decision making service. Starting with a machine learning pipeline implemented using Apache Spark, we will walk through the types of predictive models (decision trees, regressions, etc.) supported by Redis-ML, the toolkit available to load Spark models into Redis, and finally how to implement real-time decision making.