TDML Example

Test Driven Machine Learning Example

Companion exercise to the "Clean Machine Learning Code" book: http://cleanmachinelearningcode.com/

Setup

1. Requirements:

   • Docker
   • Make
   • Lots of bravery

2. Run the testable_titanic_ml_pipeline:

   • make run_with_tests_titanic_pipeline

3. Run the no_tests_titanic_ml_pipeline:

   • make run_no_tests_titanic_pipeline

Your Task

1. We received urgent new passenger data that we need to score to predict their survival:

   • datasets/titanic/2_unknown_new_data_to_score.csv

2. Unfortunately both pipelines are breaking when fed the new dataset.

3. Your task is to try to modify the no_tests_titanic_ml_pipeline.py code to support the new data and potentially save lives. Challenge yourself and set a 30 minute to get the no_tests code ready.

   • make run_no_tests_titanic_pipeline_with_new_data

4. Once you give up on the no_tests challenge because of all the inter-dependencies in the code and the lack of tests, try to do the same on the testable_titanic_ml_pipeline.py code.

   • make run_with_tests_titanic_pipeline_with_new_data

5. Send your solution in a pull request if you wish.