25: Machine Learning - Concepts

When ordinary people hear the words “Data Science,” I’ll bet the first images that come to mind are of the closely-related fields of data mining and machine learning (ML), even if they don’t know those terms. After all, this is where all the sexy tech is, and the success stories too: Netflix magically knowing which movies you’ll like, grocery chains using data from loyalty cards to optimally place products; the Oakland A’s scouring minor league stats to build a champion team with chump change (see: Moneyball). There are also creepier applications of this technology: Google placing personalized eye-catching ads in front of you using data they mined from your email text, or Cambridge Analytica projecting from voter personalities to the best ways to micro-target them.

All these examples have one thing in common: they actually make the discoveries and predictions from the data. They’re the coup de grâce. They take place after we’ve already acquired our data, imported it to an analysis environment (like Python), stored it in the appropriate data structures (like associative arrays or tables), recoded/transformed/pre-processed it as necessary, and explored it enough to know what we want to ask. All that stuff was mere prep work. This chapter is where we begin to really rock-and-roll.

---