Predictive_Modeling_and_XAI_for_GGBS_Geopolymer_Co

================================================================================ MACHINE LEARNING ANALYSIS PIPELINE: ML DATA ================================================================================ 1. OVERVIEW -------------------------------------------------------------------------------- This archive contains the experimental dataset and end-to-end Machine Learning Jupyter Notebook for predictive modeling and analysis. It serves as an open-access, reproducible computational record supporting research on material characterization and compressive strength prediction. * Dataset Source: Kaggle (rashmiverma80/ml-data) URL: https://www.kaggle.com/datasets/rashmiverma80/ml-data * Domain: Civil Engineering / Material Science / Applied Machine Learning 2. ARCHIVE CONTENTS -------------------------------------------------------------------------------- . |-- ml_analysis_notebook.ipynb # Primary Jupyter Notebook (EDA, ML models, XAI) |-- ML_data.xlsx # Primary dataset from Kaggle (or ML_data.csv) |-- requirements.txt # Python dependencies |-- outputs/ # Generated figures, models, and evaluation logs | |-- feature_importance.png | `-- prediction_scatter.png |-- LICENSE.txt # Open source and data licensing terms `-- README.txt # This documentation file 3. DATASET DESCRIPTION & ATTRIBUTES -------------------------------------------------------------------------------- * Source File: ML_data.xlsx / ML_data.csv * Origin: https://www.kaggle.com/datasets/rashmiverma80/ml-data * Primary Target Variable: Compressive Strength ($f_c'$ in $\text{MPa}$) * Handling Missing Values: Screened and addressed within Section 2 of the notebook. Feature Reference Table: -------------------------------------------------------------------------------- Variable Name | Unit | Type | Description -------------------------------------------------------------------------------- Cement | kg/m^3 | Numeric | Portland binder content Silica_Fume (SF) | % or kg/m^3 | Numeric | Mineral micro-admixture Nano_Silica (NS) | % or kg/m^3 | Numeric | Nano-engineered silica Water_to_Binder | Ratio | Numeric | Effective water/binder ratio Superplasticizer | % | Numeric | Water-reducing chemical Curing_Age | Days | Integer | Curing duration (e.g. 7, 28) Compressive_Strength | MPa | Numeric | Target strength output -------------------------------------------------------------------------------- (Note: Verify and align header names with your specific sheet columns) 4. EXECUTION INSTRUCTIONS -------------------------------------------------------------------------------- A. Running in the Cloud (Google Colab - Quickest): ------------------------------------------------- 1. Open Google Colab (https://colab.research.google.com). 2. Click 'File' > 'Upload notebook' and select 'ml_analysis_notebook.ipynb'. 3. In the left-hand files sidebar, click 'Upload to session storage' and upload 'ML_data.xlsx'. 4. Run all cells via 'Runtime' > 'Run all'. B. Running Locally (Python / Virtual Environment): -------------------------------------------------- 1. Extract the downloaded archive and open a terminal in the folder: cd ml_data_project 2. Create and activate a clean virtual environment: - On macOS/Linux: python3 -m venv venv source venv/bin/activate - On Windows: python -m venv venv venv\Scripts\activate 3. Install required packages: pip install --upgrade pip pip install -r requirements.txt 4. Start Jupyter Lab or Notebook: jupyter lab Open 'ml_analysis_notebook.ipynb' and select "Restart and Run All Cells". 5. DEPENDENCIES (requirements.txt) -------------------------------------------------------------------------------- numpy>=1.24.0 pandas>=2.0.0 openpyxl>=3.1.0 scipy>=1.10.0 scikit-learn>=1.3.0 xgboost>=1.7.0 lightgbm>=4.0.0 catboost>=1.2.0 shap>=0.42.0 matplotlib>=3.7.0 seaborn>=0.12.0 jupyterlab>=4.0.0 6. METHODOLOGY & METRICS -------------------------------------------------------------------------------- The computational pipeline implements: 1. Feature scaling via StandardScaler and RobustScaler. 2. Cross-validation: 5-fold and 10-fold stratified/k-fold CV. 3. Algorithms: Random Forest, XGBoost, CatBoost, LightGBM, and SVR/MLP. 4. Explainability: TreeSHAP analysis for global feature importance and interaction. 7. CITATION -------------------------------------------------------------------------------- If you utilize this workflow or dataset in your research, please include citations for both the repository and original Kaggle source: Dataset / Pipeline (Zenodo): [Your Last Name, First Name], and Verma, Rashmi (2026). Machine Learning Analysis Pipeline: ML Data. Zenodo. https://doi.org/10.5281/zenodo.XXXXXXX BibTeX: @dataset{zenodo_ml_data, author = {Pruthi Sagar,Neha Sharma and Verma, Rashmi}, title = {{Machine Learning Analysis Pipeline: ML Data}}, year = {2026}, publisher = {Zenodo}, doi = {10.5281/zenodo.XXXXXXX}, url = {https://doi.org/10.5281/zenodo.XXXXXXX} } Original Source Reference: @misc{verma_kaggle_ml_data, author = {Sagar Pruthi, Rashmi Verma}, title = {{ML DATA}}, year = {2024}, publisher = {Kaggle}, howpublished = {\url{https://www.kaggle.com/datasets/rashmiverma80/ml-data}} } 8. LICENSE -------------------------------------------------------------------------------- * Computational Code & Notebook: MIT License * Dataset & Documentation: Creative Commons Attribution 4.0 International (CC BY 4.0) ================================================================================