Thread by @seanjtaylor

What are the most interesting parts of building a ML classifier?
1. feature engineering
2. label engineering
3. experimenting with different models
4. hyperparameter tuning
5. summarizing/explaining model performance
6. automating training and deployment
7. <something I missed>

I have my own opinions about this that I'll share later, but I don't want to influence responses.

Great job to all the many people who pointed out parts I failed to list:
-1. Motivating and making the business/scientific case for building a classifier.
0. Collecting, cleaning, and exploring data.

Other parts that I had missed:
- studying ethical implications of the model (@profelisacelis)
- aligning model output w/ real objective (@jeremystan, @sharathrao, @rwhitcomb)
- making deployment cycle fast (@cdubhland, @vboykis)
- figuring out how to use clojure (@randyzwitch)

As promised, my thoughts on these parts:
- I liked all the answers about making sure you're solving the right problem. That is often interesting.
- I am surprised people like feature engineering and modeling so much. Machines may be taking that job away from you soon :)
[1/2]

- For me, performance evaluation is the most interesting step. Esp if you care about more than just avg perf (ties in w/ the ethics discussion).
- Automating training/deployment got little love, but useful models need to be constantly retrained/monitored and this is hard.
[2/2]