gilbertofp16/linktree_recs_apps

LINKTREE RECS APPS

In this project we will investigate and code the recommendation system to recommend top 6 apps to an Linker. I am only covering the vertical entertainment and USER_ID inside that vertical.

Repo Structure:

NOTE: Before running any of the code please download the dataset from here: https://drive.google.com/file/d/1RrlURzSB4Tb3pu8O18QI2uJZ4qt-mc0s/view?usp=share_link Then put the folder dataset in the root of the project(linktree_recs_apps/).

Research:

In this folder we will see python notebooks with the exploratory analysis. To run this code you should follow the next steps:

• execute in terminal 'curl -sSL https://install.python-poetry.org | POETRY_VERSION=1.1.15 python3 -'
• Go to research folder path and execute poetry install
• Use the new environment to run the code

Models:

In this folder we will see the training model for our recs model and then the inference module.

train_recommender_model: In this folder we can see python files with code cell. I am not using notebooks because it is more complicated to transform this code to staging or prod if I use notebook. Still this code needs better cleaning and testing.

To run the code inside this folder you should follow the next steps:

- execute in terminal 'curl -sSL https://install.python-poetry.org |  POETRY_VERSION=1.1.15 python3 -'
- Go to train_recommender_model folder path and execute poetry install
- Use the new environment to run the code

Then run the python files in the following order:
1. src/training_data.py
2. src/model_creation.py
3. src/model_evaluation.py

recommender_inference: In this folder I made the inference to use the model. This is mainly a class which will give us the recommendations. It has basic unit testing and needs more work. But it works accordingly to the requested using the function called: generate_user_recs()

To run the code inside this folder you should follow the next steps:

- execute in terminal 'curl -sSL https://install.python-poetry.org |  POETRY_VERSION=1.1.15 python3 -'
- Go to recommender_inference folder path and execute poetry install
- Use the new environment to run the code

Then test the model inference you can use the unit testing:
1. recommender_inference/tests/unit/test_model_inference.py

This module can be build and publish in poetry and then use the library everywhere. I avoid this step because
lack of time, usually we use AWS artifact in my work.