Movie recommendation system

About the project

This repository is created for educational purposes. Famous Movie Lens was considered. The version we tuned the parameters on is ml-latest.zip. The description of the dataset can be found on the website.

It is also important to understand that this repository is an attempt to present a simple solution without extremely intensive computations, which is wise considering the size of the data. Of course, the implementation is strongly coupled with the dataset, however, there are some parameters to tweak the model for a different distribution.

Additionally, the code follows the schemas of Transformer Mixin, BaseEstimator, RegressorMixin classes from scikit-learn library to be support API.

Files

Each file in the repository has it's own purpose.

File name	Meaning
.dvc	Folder for DVC
data	Directory with dataset metadata
.dvcignore	File to ensure proper data version control
movie_lens_libs.py	The library itself
analysis.ipynb	Analysis of the data and the model
sample.ipynb	A set of use cases from the library
test.ipynb	The comparison of the model with the baselines using train-test split and MSE, MAE and accuracy as metrics
visualization.py	A collection of functions responsible for visualizations

Getting started

Prerequisites

The library alone uses only Pandas and scikit-learn.

As for the date of writing this description, the following command is used to download the dependencies:

• if you are using conda:

conda install pandas scikit-learn

• if you are using pip:

pip install pandas scikit-learn

However, you might also want to play around with the analysis. Here are additional packages for data visualization and Jupyter Notebook support:

# conda
conda install ipykernel matplotlib seaborn

# pip
pip install ipykernel matplotlib seaborn

Installation

If you have all those dependencies described above, the usage is simply "plug and play".

You just have to clone the repository:

git clone https://github.com/vasia-korz/recommendation-system

Then, you can use the files within the repository. As for the usage of the library, you just have to copy movie_lens_lib.py to whatever directory you are interested in and then import in the project using Python language, for example:

import movie_lens_lib

or any other appropriate way the language allows.