Supervised Machine Learning Techniques Applied to the Core Engineering Domain

This folder contains a Jupyter Notebook that demonstrates supervised machine learning techniques applied to a materials informatics problem (predicting whether a material is metallic using the Matbench is_metal dataset).

What’s included

• supervised-learning.ipynb — full notebook with data loading, feature engineering (MAGPIE-style features via matminer), model training (Random Forest, Gradient Boosting, HistGradientBoosting, SVM, SGD), hyperparameter tuning with GridSearchCV, and model interpretation using shap.

  

Highlights

• Simple data preprocessing and feature engineering using matminer and pymatgen.
• Comparison of tree-based vs. distance-based models (scaling notes and when to apply StandardScaler).
• Hyperparameter tuning (grid search) and evaluation (accuracy, F1, ROC AUC, confusion matrix).
• Model interpretability using SHAP values and summary plots — shows which elemental features drive predictions.
• Clear, reproducible notebook that walks through analysis steps and visualizations.

Key dependencies

• numpy, pandas
• scikit-learn
• matplotlib, seaborn
• jupyterlab / notebook
• matminer
• pymatgen
• shap

Notes & dataset

• The notebook uses the Matbench matbench_expt_is_metal dataset (loaded via matminer.datasets.load_dataset). The notebook creates a composition_obj column and applies MAGPIE element featurizers which may take several minutes depending on your machine.
• Tree-based models do not require scaling; kernels and distance-based methods (SVM, KNN) do — the notebook demonstrates both cases.

Contact / Attribution

• Author: (Suyog Waghmode)

Linked files

• requirements.txt — environment packages list next to the notebook for reproducibility.

---