Course workload analytics

Supplementary repository including data preprocessing, machine learning training, and data analysis code for manuscript "Insights into undergraduate pathways using course load analytics" accepted as full paper to LAK '23.

Citation:

Borchers, C., & Pardos, Z. A. (2023). Insights into undergraduate pathways using course load analytics. In Proceedings of the 13th International Learning Analytics and Knowledge Conference (LAK). ACM. https://doi.org/10.1145/3576050.3576081

@inproceedings{borchers2023insights,
  title={Insights into undergraduate pathways using course load analytics},
  author={Borchers, Conrad and Pardos, Zachary A},
  booktitle={LAK23: 13th International Learning Analytics and Knowledge Conference},
  doi={10.1145/3576050.3576081},
  url={https://doi.org/10.1145/3576050.3576081},
  year={2023}
}

Folder structure

• utils.py: Collection of functions and classes used throughout scripts.
• 1-create-course-features.py: Script to create LMS and enrollment-based variables for individual semesters and save results to .csv files. Run via python3 1-create-course-features.py '2021 Spring' or any other semester between Spring '17 and Spring '21.
• 2-prepare-training-data.py: Script to pre-process survey data of Spring '21 semester and join course-level variables created via 1-create-course-features.py to the survey data.
• 3-train-model.py: Script to perform cross-validated random search for a set of models specified in run_model_training() and a specified set of target variables (LABELS) in utils.py. All relevant results are saved as pickle objects to the models/ folder.
• 4-evaluate-models.ipynb: Notebook to export the cross-validation error and test error and perform robustness checks for predictions with high discrepancy to credit hour designation.
• 5-extrapolate.ipynb: Notebook to cread course-level course load predictions for all previous semesters based on LMS and enrollment data for a specified set of models and target variables.
• 6-analysis.ipynb: Notebook to perform descriptive and inferential analyses on the scaled course load predictions on a course and student-semester level. Includes code to reproduce plots in manuscript.
• 7-modeling.R: R code to perform likelihood-ratio tests for model selection on logistic regression models of dropout and on-time graduation.