Cuisine Predictor

Developer: Biswas Nandamuri

Cuisine Predictor is a python based tool wrapping ML models which use Scikit Learn's LinearSVC and kNeighborsClassifier to predict the cuisine and similar dishes from Yummly catlog.

The project's python code follows PEP8 Style Guide

Dependencies

• Scikit-learn - scikit-learn is a Python module for machine learning built on top of SciPy
• Pandas - Flexible and powerful data analysis / manipulation library for Python
• Nltk - NLTK is a platform for building Python programs to work with human language data

Dev Dependencies

• Pytest - Testing framework that supports complex functional testing
• Pytest-cov - Coverage plugin for pytest
• autopep8 - Tool that automatically formats Python code to conform to the PEP 8 style guide
• Jupyterlab - Browser-based computational environment for python
• Matplotlib - Matplotlib is a comprehensive library for creating static, animated, and interactive visualizations in Python
• Seaborn - Seaborn is a Python data visualization library based on matplotlib

Run on local system

1. Clone this repository and move into the folder.

   $ git clone https://github.com/Biswas-N/cs5293sp22-project2.git
   $ cd cs5293sp22-project2

2. Install dependencies using Pipenv.

   $ pipenv install

3. Run the utility tool

   $ make

   Note: Project includes a Makefile which has commonly used commands. By running make the following command pipenv run python project2.py --N 5 --ingredient "chili powder" --ingredient "crushed red pepper flakes" --ingredient "garlic powder" --ingredient "sea salt" --ingredient "ground cumin" --ingredient "onion powder" --ingredient "dried oregano" --ingredient paprika is executed.

   Note on Model: If pre-fitted models does not exist in models folder, this tool creates fitted-models based on assets/yummly.json data and stores them in models folder using joblib. So the first execution may take more time than typical execution times.

Documentation

The documentation about code structure, model building and prediction algorithm can be found here.

Testing

This utility is tested using pytest.

Documentation about the tests can be found here. Follow the below commands to run tests on your local system.

1. Install dev-dependencies.

   $ pipenv install --dev

2. Run tests using Makefile.

   $ make test

3. Run test coverage.

   $ make cov

Bugs/Assumptions

• This tool assumes that the yummly.json is present in assets folder. So if the data file is not present in the folder, this tool may fail.
• Similarly, this tool assumes there are three keys in each JSON object in yummly.json file called id, cuisine and ingredients (should be a list of ingredients). If the data inside yummly.json is not as expected, the tool may fail.
• This tool is built using LinearSVC and KNeighborsClassifier, and trained using the given yummly.json data. So this tools accuracy is based on the data quality provided and statistical techniques used in the above said models. I tried using the best approaches possible for pre-processing and model fitting in the given time constraints. But there is always more to do, so there might be cases in which the tool can predict less-accurate results.