These experiments are examples I discuss during a talk of this title.
More than machine learning is necessary for delivering data science solutions. The other stuff includes cleaning data, exploring data, engineering features, and operationalizing your model. All of these require **knowing your data, your domain, your question, and how you will use the answer**. There are four illustrative experiments: 1. Exploration by visualization 2. Handling missing values 3. Feature engineering for trains 4. Publishing an API for trains Slides and video that walk though the example are in [the blog post] . I am a Senior Data Scientist at Microsoft. : http://brohrer.github.io/data_science_other_stuff.html