/TransferRecommendationChildren

Analysis conducted to investigate the usefulness of transfer learning when designing recommendation systems for children.

Primary LanguageHTML

This archive contains the scripts to replicate the experiment for our paper "Can we leverage rating patterns from traditional users to enhance recommendations for children?" published in ACM RecSys by - Ion Madrazo Azpiazu, Michael Green, Oghenemaro Anuyah, and Maria Soledad Pera.

Requirements

  • Java
  • An R, Jupyter, Python, and Tidyverse installation.
  • The MovieLens 1M dataset, extracted into Data/input (you should have directory Data/ml-1m)
  • Any dataset containing ratings provided by children extracted into Data/input

Instructions

  • Steps to run:

    • Install required software and data files enlisted in requirements. These directories and files should be present upon doing so: -Data/input/ml-1m (e.g. data/ml-1m/ratings.dat) -Data/input/childrens_data (e.g., any children ratings file)

  • Run Jupyter notebook:

    • Create_Experimental_Datasets/Data_creation_notebook.ipynb

  • Run LibReC Experiment:

    • Input_Analysis/

  • Visualize user-rating activity:

    • Rating_Distribution_in_datasets.ipynb