AI-Powered Business Intelligence

Chapter 7. AI-Powered Predictive Analytics

Get ready for takeoff! We are entering the space of AI-powered predictive analytics. And we will do that by stepping into the role of a BI analyst at American Airlines. In this chapter, we will look at three use cases from a real-world dataset. First, we will try to classify flights as to whether they will land on time or not. Second, we want to detect bottlenecks in our flight schedule by forecasting actual flight times in contrast to the scheduled flight duration. And finally, we will analyze airports’ ability to keep up with the flight schedule by using automatic anomaly detection.

Our goal is to build a prototype of an AI-powered BI solution that proves to solve a specific problem depending on the use case. We want to evaluate how well the AI model works with our data, show it around, and gain support for the new approach in our organization by demonstrating the business value (build fast, show fast, learn fast), as outlined in Chapter 4. To that end, we will start by abstracting away things like setting up data pipelines, handling ETL jobs, and integrating AI services into our enterprise data warehouse. But rest assured—all of these things will be possible, as you will learn in Chapter 11.

What will most likely be the same in the prototyping and production phases is the AI model we will build. Already in our prototype, we will use enterprise-grade AI services from Microsoft Azure that will stand up to production workloads. The only ...