Machine learning (ML) has been making great strides due to its successful adoption into various disciplines. This has fueled a growing demand for ML experts (aka ‘Data Scientists’) who are short in supply. Numerous easy to use software applications (like Sci-kit learn, H20, Keras, Weka) have been developed over the last decade to bridge this gap. Building good models with these tools still needs reasonable knowledge of the algorithms. Our inherent human desire to automate paired with a shortage of ML experts led to automation of the modeling pipeline – “Automated Machine Learning” (AutoML) tools.
This paper focuses on the “state” of AutoML. Part 1 includes a high-level overview of AutoML, its history, its place in data science, and why data scientists will ultimately embrace AutoML as a key tool in their arsenal of data mining capabilities. Part 2 reviews the level of sophistication currently available in AutoML by evaluating available technology in this space.