FEUP-DISS

This repository hosts the Software library on stream based recommender systems. This is part of my thesis dissertation done at the Faculty of Engineering of the University of Porto.

General Overview

Contents

This software library contains code regarding the implementation of recommendation systems on online platforms. It implements some algorithms, test metrics and data structures, so that it can process incoming information, denoted as a stream of ratings. Each rating can be implicit or explicit and is given from an user to an item. Explicit ratings are direct evaluations given by users, for example on a 1-10 scale. Implicit ratings can be interpreted as actions users took on items(for example a page click) which can be interpreted as the user liking that specific item. Each rating is a tuple : (user_id, item_id, rating). For implicit feedback the tuple becomes: (user_id, item_id)

Current implemented algorithms:

1. Explicit User-Based Collaborative Filtering
2. Explicit User-Based Clustering Collaborative Filtering
3. Implicit User-Based Collaborative Filtering
4. Implicit User-Based Clustering Collaborative Filtering
5. Implicit Locality-Sensitive-Hashing Item-Based Collaborative Filtering
6. Implicit Locality-Sensitive-Hashing User-Based Collaborative Filtering
7. Implicit Matrix Factorization
8. Explicit Matrix Factorization with matrix preprocessing
9. Explicit Matrix Factorization without matrix preprocessing
10. Implicit Item-Based Collaborative Filtering
11. Implicit Item-Based Clustering Collaborative Filtering

Install

Currently the library is deployed on Pypi.

pip install increc

Dependencies

For development purposes:

1. Python3.8
2. Pipenv is required to run this on a virtual environment. pip install --user pipenv
3. python -m pipenv lock -r > requirements.txt
4. pipenv install -r requirements.txt

Run Tests

This library uses unittest for unit testing and flake8 as its linter.

1. pipenv run python -m unittest discover -v -p "*_test.py"
2. pipenv run flake8

Run Examples

Usage examples are located in the examples folder.

pipenv run python -m examples.collaborative_filtering.neighborhood.explicit_feedback.user_based.user_based_clustering "data_set"

Documentation

Documentation is created using Sphinx.

1. pipenv run sphinx-apidoc -f -o docs/source algorithms
2. pipenv run sphinx-apidoc -f -o docs/source stream
3. pipenv run sphinx-apidoc -f -o docs/source data_structures
4. pipenv run sphinx-apidoc -f -o docs/source utils
5. pipenv run sphinx-apidoc -f -o docs/source graphic
6. pipenv run sphinx-build -b html docs/source/ docs/build

https://marko50.github.io/increc-documentation