Deploying a model for batch or real-time predictions
You can train a model anywhere, so why bother using Azure ML? Mostly because it gives us a very easy way to deploy the model to integrate it with Power BI. Before we get into how to deploy the model with Azure ML, let's go over some theory behind why we would want to deploy a model at all.
Training a model can cost a lot of compute power. The reason why AI has had a resurgence is due to the easy availability of cloud compute, among other reasons. Instead of needing a computer with enough processing power, we can now simply use whatever compute is necessary from any cloud provider to train the model, and not use it when we're done training the model.
By having more compute at our disposal, we can train models with complicated algorithms, but also use more data, which generally improves a model's accuracy.
However, training a model and using that trained model to generate new predictions, doesn't require...
