shriman/attraqt_movie_recommendation

Attraqt movie recommendation

This is the submission code for movies recommender test.

Please read the documentation below to run the code and generate the results.

Colab notebooks and preliminary data explorations

The repository contains two colab notebooks, which can be viewed either here in GitHub or in Google Colab (links are in files.)

• Attraqt.ipynb :

  • Initial data exploration and data cleaning.
  • Basic recommender: Recommender based solely on movie ratings. Predicts top n movies of all time.
  • Genre-based movie recommender (top-K movies): Recommends top-n movies in each genre.
  • Metadata-based recommender / Content-based filtering Recommends movies based on tagline and overview, using tfidf on text.

• Attraqt_collaborative_filtering_svd.ipynb:

  • Collaborative filtering based recommender using the user ratings.
  • Submission: code to generate the submission results.
  • Advanced recommender: using a combination of collaborative filtering and content-based filtering approach.

Set-up

• Clone the repository into your local folder.
• To run the code, you will need to have Python 3.9 and pip installed.
• Run the make command to create your virtual environemnt: make
• Activate your virtual environment source venv/bin/activate

Intial configuration and Data files

• The file configs/config.yml contains path to all the source files as well output files
• The data files are located in the folder data in the project path, where project path is the same as the directory of this README file.
• Please note that the config file has three sections:
  • data : path to data files (movies_metadata.csv, ratings.csv, evaluation_ratings.csv, and true_ratings.csv)
  • output: path to output files
  • model: path to trained model

data:
  movies_metadata: 'data/movies_metadata.csv'
  ratings: 'data/ratings.csv'
  evaluation_ratings: 'data/evaluation_ratings.csv'
  true_ratings: 'data/true_ratings.csv'

output:
  submission: 'data/outputs/submission.csv'
  bonus_submission: 'data/outputs/bonus_submission.csv'

model:
  trained_model_path: 'data/outputs/svd_trained'

• It is assumed that the true ratings are stored in the file true_ratings.csv in the path data/true_ratings.csv and has the following column names: UserId, ModelId, TrueRating

Usage

The project used cli to run the commands

• Train and save the model on the disk

# train the model and save to disk
$ python -m attraqt train

• Predict the ratings on evaluation dataset and generate the submissions file

# predict the ratings on the evaluation dataset
# and save results to submission.csv
$ python -m attraqt predict

• Evaluate the results using RMSE

# predict the ratings on the evaluation dataset
# and save results to submission.csv
$ python -m attraqt evaluate

• To correctly run the evaluation, it is assumed that the true ratings are stored in the file true_ratings.csv in the path data/true_ratings.csv and have the following column names: UserId, ModelId, TrueRating

• The trained model is can be downloaded from here

• The initial submissions file can be downloaded from here