This repo holds my weekly work for the AnalystLab Africa Data Science Internship. Weeks 1–3 used one dataset (Loan Prediction, after Week 1's separate HR attrition warm-up). From Week 4 onward, the program moved into the HealthConnect Experience Lab — a multi-track group project (Data Science, Data Analytics, ML Engineering, Generative AI, Project Management) built up week by week through problem definition, baseline modelling, integration, testing, and final presentation.
- Week 1 — Employee attrition exploration (IBM HR dataset). A separate warm-up project before settling on the Loan Prediction dataset used from Week 2 onward.
- Week 2 — Cleaning and preparing the Loan Prediction dataset for machine learning.
- Week 3 — Statistical testing and feature engineering on top of the Week 2 cleaned dataset.
- Week 4 — HealthConnect Experience Lab kickoff: problem definition for predicting appointment no-shows, target and feature-exclusion decisions.
- Week 5 — Baseline no-show model: data preparation, feature engineering, and a first logistic regression baseline (ROC-AUC 0.677).
- Week 6 — Error analysis on the baseline, feature refinement, and model comparison — Random Forest candidate (ROC-AUC 0.686).
- Week 7 — Systematic testing of the candidate model: overfitting check, cross-validation, business-relevance testing, pipeline robustness.
- Week 8 — Final model documentation, ML Engineering handoff, and project presentation materials.
python -m venv venv
venv\Scripts\Activate.ps1 # Windows
pip install -r requirements.txt
Each week's notebook lives under its own folder — open the relevant one and select the venv kernel.
Caleb Djarabé — Data Science Intern, AnalystLab Africa