Introduction

Creating predictive models is obviously very useful. But predictive models are even more useful when they are "built" into an information system that can use the prediction to optimize consumer or user behaviors. For example, hospitals may create predictive models to help them estimate how many hours of nurse care are required for each patient who comes into the emergency room (ER). However, in order for those predictions to be truly useful, the patient registration system needs to use the prediction to estimate total nurse staffing needs as patients come into the ER. Furthermore, new data on the actual nursing hours required for each patient needs to be continually recorded into the same system so that the predictive models can be continually and automatically retrained.

Integrating predictive modeling into information systems is what makes them "intelligent", or as it has become popularly referred to as, "machine learning". This supplemental chapters will give you a few simple examples on how to call the Azure ML Studio APIs from external apps and how to retrain predictive models dynamically. Thus, is teaches you how to create a rudimentary machine learning environment.

At a minimum, a machine learning environment will require the following:

  1. Development of a predictive model based on prior behaviors and data in order to improve the future behaviors and decisions of consumers or end users

  2. Dissemination of that prediction to the consumer/end user

  3. Storage of the new/future behaviors and data that were used to create that prediction

  4. Transfer of those new behaviors to the analytical database (i.e. ETL)

  5. Automated retraining of the predictive model for improvement of future predictions

You've spent most of your time in this book already learning Step 1. Step 2: Disseminating the model, typically requires building an app of some type that will use the predication. Azure ML Studio makes that step very easy by auto-generating an API that can be called from almost any software architecture. The following section will give you an example of Step 2. Then, the next section will combine Steps 3 and 4 by storing behaviors in the same database that the predictive model is built upon using a product recommendation engine example. Finally, the last section will show you how to automatically retrain your Azure ML Studio model.