In-class activity 8: Support vector machines, part 2

There is no starter code this time. You'll be doing everything from scratch. Start a Jupyter notebook with

jupyter notebook --ip=127.0.0.1 &

Your task is

• Go to Kaggle (or another source of data sets, if you have another one that you've used) and find a dataset, one that you and your partner have not worked on before. Spend some some time studying the data. The "Look at the data" section from Lab 1 may give you some ideas on how to do this.
• Formulate a classification question.
• Clean up the data to get it into a state that you can use for this question.
• Split the data into a training and test set.
• Train an SVM classifier on the test set.
• Evaluate the classifier.
• Go back to a previous step to see if you can make a more accurate classifier. It's possible that you need to go all the way back to the "formulate a question" step, because maybe there is a slightly different, more useful question that you can address more accurately.

Make every effort to do something new and interesting.

As you go along, document what you do---including (to a certain extent) things you try but don't work---in the Jupyter notebook, because you will be turning this in as a summary of what you did in this activity. Include verbiage as you go along explaining what your data set it, what quesiton you're asking, rationale for how you clean it up, etc. At the end write a paragraph or so about how well your classifier does and what conclusions you draw from that.

Make sure you save save your notebook at the end---Jupyter does do some autosaving, but not continually (as, for example, Google Docs and Sheets does). Give your notebook at name something like

lab8-svn2-STUDENT1NAME-STUDENT2NAME

That is, put both of your names in the title. Then copy it into one of your turn-in folders.

The most important documentation for today is the API for sklearn's SVM classes, but here are links to the other libraries, as needed: