AutoML is the concept of automating the practical aspects of machine learning model development - the creation of models, hyper-parameter tuning, algorithm selection, and feature management. The degree of automation can vary widely - from manual control of algorithm selection and tuning to a fully automated "one-click" solution. It simplifies the process of training and evaluating machine learning models so that data scientists can focus on business problems.
Practical machine learning requires many repetitive tasks. Data scientists explore data and transform it in different ways before even choosing to begin machine learning. Machine learning then requires many iterations of tuning and testing before data scientists are confident in a model. This means that an onerous amount of time is spent on repetitive, mundane tasks and on working with code. AutoML simplifies this using a workbench environment that automates all technical aspects of Machine Learning. Data scientists can now spend time working on domain problems and model accuracy than writing code.
