How hammilidi/diabete_prediction works
Project type and purpose
• This is a data-science/ML pipeline project whose goal is to cluster patients by diabetes risk and then train a logistic-regression classifier (plus comparisons) to predict those clusters.
• The README confirms it’s a Python-based ML workflow: exploratory data analysis (EDA), preprocessing, clustering, model training with hyperparameter tuning (GridSearchCV), evaluation, and model persistence.
File structure
• Top-level directories:
• data/ – raw CSV dataset
• scripts/ – Jupyter notebooks and a preprocessing module
• models/ – serialized model artifact
• scripts/ contains:
• analyse_exploratoire.ipynb (EDA & visualization)
• preprocessing.py (data-cleaning, feature-engineering, cluster-creation functions)
• cluster.ipynb (cluster analysis)
• Models.ipynb (training + hyperparameter optimization)
• Test_models.ipynb (final evaluation)
• requirements.txt lists core libraries; no explicit deployment or service definitions.
README for additional insights
• Confirms use of pandas/numpy (data handling), scikit-learn & imblearn (modeling/pipelines), matplotlib/seaborn (plots).
• Describes methodology: outlier cleaning, normalization, cluster determination, GridSearchCV, cross-validation, model comparison, selection, serialization.
Key reminders
• Every stage—even if it’s a Jupyter notebook—as a distinct component
• Iterative loops (e.g., hyperparameter search) explicitly
• Labels map directly to files and functions in the repo for traceability