Recommendation Systems

This material is based on this Alura course.

The material from course is implemented using Jupyter Notebooks, but here I implement a variation using only python and a different project structure. The goal is to implement my own version of KNN to recommend movies to a user.

The data-source can be obtained here. The small dataset is included on this repository, but the full dataset must be downloaded separately.

Dependencies

• pandas

• numpy

• fastAPI

  pip install pandas 
  pip install numpy 
  pip install fastapi
  pip install "uvicorn[standard]"

Running de Code

There is two ways to run the code: (1) running the recommendations on the console and (2) using a FastAPI server.

(1)

python main.py

(2)

uvicorn main:app --reload

The reload option is used to reload the code when it changes and should be used only when developing.

For te API, the Documentation can be found at http://localhost:8000/docs.

Improvements

• Start using mongodb to store the data and not in a csv file.
• Create workers to preprocess de data.
• Training with the complete dataset.