Integrating machine learning

When to use machine learning

If your company has a large set of data of past decisions, data scientists can use this data to create a machine learning model. This model can then predict the outcome of new decisions based on this data. The accuracy of the prediction can vary depending on the size and range of the data set.

When or after you design and implement a decision model, you can enrich it by combining rules that describe decisions with machine learning that make predictions.

How to use machine learning in a decision model

Data scientists deploy machine learning models on a machine learning platform, such as Watson™ Machine Learning. Then, you must configure the access from a decision solution to the machine learning providers that contain the model deployments. These providers become available in Decision Designer and you can import deployments or serialized models from these providers to your decision services.

Alternatively, data scientists can also provide you with transparent machine learning models that can be imported into Decision Designer without any prior configuration of the platform.

When you import a machine learning model, it generates a predictive model template that contains all the elements to invoke the machine learning model. You complete this template so that the predictive model can consume your machine learning model.

Finally, you encapsulate the predictive model in a decision node in your decision model. When the decision model is executed, the machine learning model computes a prediction based on the inputs of the decision node that contains it.