Pytest for Unit Tests and TravisCI for Continuous Integration
Run tests with cmd: pytest in terminal from production-ml folder.
I used Python and pytest to build a Production Machine Learning App that loads data, preprocesses it, and provides preliminary EDA (Exploratory Data Analysis). The app currently supports structured data, and it is extensible to unstructured data (text and images).
- Used Pytest to test Matplotlib's plotting, had to override object's equal function and configure Travis's VM to make this successful
- Made EDA object Immutable: analytic functions return new EDA objects on the fly with different configurations
- Extensibility: Currently the project supports Structured Data, but the app architecture is extendable to Unstructured Data as well. In particular, text and images will also have their own preprocess and EDA objects.
- Made use of delegation: The Dataset object delegates to Preprocess, which delegates to EDA
- StructuredDataApp defines Dataset, Features, Model, Train, and Eval objects
- StructuredDataApp inherits from MachineLearningApp
- MachineLearningApp initializes Dataset, Features, Model, Train, and Eval objects
- A future class "UnstructuredDataApp" will inherit from MachineLearningApp as well
- Its Dataset object utilize different methods for preprocessing text and image data