Analysis of Student Behavior and Score Prediction in ASSISTments online learning

Authors: Aswani Yaramala, Soheila Farokhi, and Hamid Karimi

The initial dataset files are located in the data/ folder.

Code Setup and Requirements

You can install all the required packages using the following command:

    $ pip install -r requirements.txt

Code for Initial Data Exploration

Code for dataset exploration and creating the plots in the Section 3.2 of the paper, is in the Initial_data_statistics.ipynb notebook file.

Code for Tutoring Request and Student Performance

Code for analysis in Section 4.1 of the paper, is in the Tutoring_analysis.ipynb notebook file.

Code for CCSS Skill Mastery and Student Performance

Code for analysis in Section 4.2 of the paper, is in the AssociationRule_skills.ipynb notebook file.

Code for Feature Extraction

Code for extracting hand-crafted features detailed in Section 5.1. of the paper, is in the Feature_engineering.ipynb notebook file. This creates the dataset in setting (I). The files are saved in saved_files/ directory. This step is necessary for creating other dataset settings and training predictive models on the dataset.

Creating the Graph

To create the dataset in settings (II) and (III), use the following command. This will save the edge list for the graph in saved_files/ directory a model for the learned embeddings in models/ directory. Also, the final train and evaluation dataset for setting (II) or (III) will be saved in the saved_files/ directory.

   $ python graph_dataset.py --setting <dataset_setting>

--setting is the setting of the dataset described in the paper. The value for this parameter should be either 2 for setting (II) or 3 for setting (III).

Hyperparameter Tuning for Predictive Models and Feature Importance

To run hyperparameter tuning for 5 predictive models on the dataset in setting (I), run the code in Predictive_models_Feat_importance.ipynb. This notebook also contains code for creating the feature importance plot in Section 5.5.1.

To run hyperparameter tuning for 5 predictive models on the dataset in settings (II) or (III), use the following command. This will save the models for classifiers in the models/ directory.

   $ python tune_models_with_embedding.py --setting <dataset_setting>

--setting is the setting of the dataset described in the paper. The value for this parameter should be either 2 for setting (II) or 3 for setting (III).

References

EDM Cup 2023

@misc{Prihar_Heffernan_2023,
  title={EDM Cup 2023},
  url={osf.io/yrwuh},
  DOI={10.17605/OSF.IO/YRWUH},
  publisher={OSF},
  author={Prihar, Ethan and Heffernan, Neil T, III},
  year={2023},
  month={Jun}
}