Model Card for Model ID

This model recommends movies to users based on the movies they have voted for.

Model Details

The model consists of three parts

Content-based filtering
Collaborative filtering
Ensemble model

Model Description

Content-based filtering is used for recommending movies based on the content of their previously voted movies. e.g. genre, actors, ...

By using collaborative filtering, similar interests are found and movies that have been voted for by some users are recommended to users who have not voted for them. It doesn't depend on the content and doesn't need domain knowledge.

The ensemble model is created by combining the last two methods to give better recommendations. The algorithm finds similar people and then recommends films based on their votes, filtering them based on content preferences.

Developed by: Aida Aminian, Mohammadreza Mohammadzadeh Asl

Model type: content-based filtering and collaborative and an ensemble model of these two model
Language(s) (NLP): not used, only TFIDF for keywords is used
License: MIT License

Model Sources

MovieLens dataset

Uses

Building recommendation systems

Direct Use

Movie recommendations based on content and other similar people.

Bias, Risks, and Limitations

This ML model is based on an IMDB movie dataset. The dataset may have more focus on English movies.

Recommendations

Add other metrics to model

How to Get Started with the Model

Install the sklearn, pandas and numpy libraries for python. Download the MovieLens dataset and put that in the 'content/IMDB' path in the project directory. Use python interpreter to run the code.

Training Details

Training Data

IMDB Movies

Preprocessing

Extracting features from keywords.

Evaluation

Factors

We've removed some of the wrong rows in the dataset.

Metrics

percision@k and recall@k

Summary

Technical Specifications

Model Architecture and Objective

Content-based filtering.

Hardware

Works fine on google colab

Software

python, sklearn, numpy, pandas