Implement the K-Means Classifier

Tasks

Create a new branch in your data structures repository called kmeans. You’ll use this branch to do your work for this assignment. Add a new file to your data-structures repository called k_means.py. In this file, implement a K-Means Classifier object.

On initialization, your K-Means Classifier should take semi-optional parameters max_iter and min_step. If one is specified you don’t need the other.

Your classifier should have the following public methods:

• clf.fit(self, data, k): Generates k centroids with which to classify the given data. Does not return anything
• clf.predict(self, data): Returns classes for some data if clf.fit has already been called. This should be able to take in an iterable of data records (either a tuple of records, or a list of records, or Pandas DataFrame). If clf.fit has not yet been called, raise an appropriate error.

The classifier should default to 2 centroids. We’ll also only work with two characteristics for this, though you can generalize your classifier if you want. Appropriately handle the case where k is provided with an invalid input (e.g. -1, or k > len(data))

Your implementation must include tests. The tests must run in both Python 2 and Python 3. Ensure that your tests are properly connected to Travis CI and that you have a Travis badge displayed on your README.md. Expand your README.md to include notes about your classifier. Include any external sources used in its creation.

:download:`Click here </downloads/flowers_data.csv>`_ for a data file of classified flowers which you may use to test your classifier. You can either use the column named “target” or “class_name” as your class label. They each map directly to each other.

You can also use the :download:`lean_stars_and_galaxies.csv </downloads/lean_stars_and_galaxies.csv>`_ data shown in the notes. If you do, ask for help.

Include the time complexities for this classification algorithm’s fit and predict methods in your README.

Stretch Goal

Enable your classifier to use as many characteristics as you provide. For example, if your input data from flowers_data.csv looks like (petal_length, petal_width, sepal_length, sepal_width), your classifier should be using the total distance between points. If unclear, ask.

Submitting Your Work

When you’ve completed your work and all your tests are passing, submit a pull request from the kmeans branch back to master. Copy the URL for that pull request and submit it using the URL input on Canvas. Once submitted you may merge your branch back to master.

Use the comment function to submit quests, comments, and reflections on the work you’ve done.