This session will include a moderated Q&A featuring questions from the live audience.
Machine learning and the related fields of data science, data mining, and predictive analytics have been a key part of advanced analytics solutions for decades. The emergence of automated machine learning (AutoML) is poised to take enterprise analytics to the next level.
Historically, automation in machine learning has been limited to what is done in the model-building algorithms themselves, such as variable selection done by decision trees.
During the past decade, increasingly, machine learning software has added additional automation to the model-building process, such as hyper-parameter tuning in neural networks and model ensembles.
More recently, the term AutoML has arisen to describe the automation of as much of the machine learning process as possible. However, the primary focus has been on the algorithms. Insufficient progress has been made with the most time-consuming part of the machine learning process: data preparation.
The future of AutoML should and must include significant advances in automating data preparation steps. This talk will review the current state of AutoML and propose areas where advances in data preparation steps will be added to AutoML.