How jlgonrod/ibd_predict works
Project type and purpose
• This is a data-science / machine-learning pipeline (no web or mobile frontend) whose goal is to build and validate predictive models (XGBoost) for medium- and long-term clinical remission and response in IBD patients treated with vedolizumab or ustekinumab.
• It is driven by Jupyter notebooks, one R Markdown script, and produces model artifacts and Excel tables for reporting.
File structure
• Top-level artifacts:
• Several notebooks (“1A_generate_synthetic.ipynb”, “1_read_and_cleaning.ipynb”, “2_EII_Analytica.ipynb”, “3_EII_analityca_histogram_antitnf_age.ipynb”) that implement each stage (data generation, cleaning, exploratory analysis, histogram plotting).
• “N_ajusted.Rmd” – supplementary analysis or reporting in R.
• “output/” subtree containing subfolders:
– “models/” holds trained XGBoost model JSON/text files and fairness analysis results.
– Excel files with tables (cohort summaries, subcohort tables).
• “requirements.txt” – Python libraries.
• “README.md” – project description, clinical context, variable definitions.
README for additional insights
• Describes clinical problem, cohorts, inclusion/exclusion, independent & dependent variables.
• Targets: respuesta_clinica_26s, respuesta_clinica_52s, remision_clinica_52s.
Detailed capture of essential elements
• Four notebooks and the Rmd script as individual components.
• “output/models” and “output/*.xlsx” directories explicitly.