Artificial Intelligence with Microsoft Power BI

Chapter 10. Custom Machine Learning Models

In Chapter 3, we discussed how to train and host a custom machine learning model within Power BI using AutoML. But what if you already have a trained model from a data scientist at your organization? Or you want to specify which algorithm to use? Or you want to more closely monitor and interpret a model? In these cases, you will need another platform that can deploy a custom machine learning model and still connect with Power BI. The solution here is to use Azure Machine Learning (details later on how to get a free trial of this through Azure). We will show you how to train your own custom model in Azure and use it to score data within Power BI.

AI Business Strategy

In 2013, the MD Anderson Cancer Center in Houston, Texas, started an AI project to diagnose and recommend treatments for cancer patients. After four years of development and $62 million in costs, the project was paused before ever being used on patients. At the same time, the center’s IT department successfully implemented AI models to make hotel and restaurant recommendations for patients’ families, identify patients who needed help to pay bills, and address staffing problems.¹

The lesson from the story is that smaller-scale applications are often much more successful than moonshot projects. We do not, however, want to deter an organization from being ambitious; instead, a business should start small and scale up—in terms of both project scope and organizational capabilities. ...