🧠 FAI: Feature-Wise Adaptive Imputation

"One imputation method does not fit all features — FAI learns to choose wisely."

A machine learning framework that automatically selects the best imputation method per feature, optimizing for downstream predictive performance — not just imputation error.

📖 Overview • 🧪 Methodology • 📊 Results • 📁 Structure • 🚀 Quick Start • 👨‍💻 Author

---

📌 Overview

Missing data is one of the most common challenges in real-world machine learning. Most practitioners apply a single imputation method to all features — but this ignores the fact that different features have different statistical properties, and therefore benefit from different imputation strategies.

FAI (Feature-Wise Adaptive Imputation) solves this by:

• 🔍 Analyzing each feature's statistical characteristics
• 🧪 Evaluating multiple imputation methods via cross-validation
• 🤖 Learning a selector that maps feature properties → best method
• ⚡ Applying the optimal imputation per feature automatically

✨ Key Highlights

Feature	Description
🎯 Per-feature selection	Each feature gets its own optimal imputation method
📈 Downstream-aware	Optimizes Accuracy, F1, and AUC — not just RMSE
🤖 Fully automatic	Zero manual tuning required
🔬 Research-ready	Publication-quality figures and CSV results
⚡ Scalable	Supports MCAR, MAR, and MNAR mechanisms

---

🧪 Methodology

Raw Data with Missing Values
         │
         ▼
┌─────────────────────────┐
│  Step 1: Feature        │
│  Descriptor Computation │  ← missing rate, skewness, variance,
└────────────┬────────────┘    correlation, entropy, data type
             │
             ▼
┌─────────────────────────┐
│  Step 2: Isolated       │
│  Label Generation       │  ← 3×5-fold CV per feature per method
└────────────┬────────────┘    using downstream model performance
             │
             ▼
┌─────────────────────────┐
│  Step 3: Selector       │
│  Training               │  ← Random Forest: descriptors → method
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│  Step 4: Adaptive       │
│  Inference              │  ← predict method per feature → impute
└────────────┬────────────┘
             │
             ▼
    Imputed Dataset ✅

🔧 Imputation Methods Compared

Method	Type	Best For
Mean	Statistical	Symmetric numerical data
Median	Statistical	Skewed numerical data
KNN	ML-based	Low-dimensional data
MICE	Iterative	MAR data with correlations
missForest	Ensemble	Complex non-linear patterns

---

📊 Results

Performance Summary — Adult Dataset

Mechanism	Missing Rate	FAI Accuracy	Best Baseline	Gap
MCAR	5%	0.8639	0.8656 (MICE)	-0.17%
MCAR	10%	0.8586	0.8633 (Mean)	-0.47%
MCAR	15%	0.8447	0.8512 (MICE)	-0.65%
MAR	5%	0.8520	0.8578 (KNN)	-0.58%
MAR	10%	0.8443	0.8452 (MICE)	-0.09%
MAR	15%	0.8325	0.8355 (MICE)	-0.30%
MNAR	5%	0.8463	0.8486 (Mean)	-0.23%
MNAR	10%	0.8262	0.8268 (Mean)	-0.06%
MNAR	15%	0.8179	0.8219 (MICE)	-0.40%

✅ FAI is consistently within 1% of the best baseline — fully automatically, with zero manual tuning.

📈 Visualizations

Accuracy Comparison	F1 Score Comparison

AUC Comparison	Method Selection Heatmap

Feature Descriptor Importance

---

📁 Repository Structure

FAI-Feature-Wise-Adaptive-Imputation/
│
├── 📓 A_survey_on_missing_data_in_machine_learning.ipynb
│       └── Complete Google Colab notebook (all cells)
│
├── 🖼️  Images/
│       ├── figure_accuracy_comparison.png
│       ├── figure_accuracy_comparison_bar.png
│       ├── figure_f1_comparison.png
│       ├── figure_f1_comparison_bar.png
│       ├── figure_auc_comparison.png
│       ├── figure_auc_comparison_bar.png
│       ├── figure_method_selection_heatmap.png
│       ├── figure_method_selection_stacked.png
│       ├── figure_descriptor_importance.png
│       └── figure_descriptor_importance_vertical.png
│
├── 📊 CSV_Files/
│       ├── fai_combined_metrics.csv
│       ├── fai_compact_improvement.csv
│       ├── fai_detailed_improvement.csv
│       ├── fai_experiment_results.csv
│       ├── fai_improvement_summary.csv
│       └── fai_paper_results_table.csv
│
├── 📄 requirements.txt
└── 📖 README.md

📊 CSV Result Files

File	Description
fai_combined_metrics.csv	Accuracy, F1, AUC across all scenarios
fai_compact_improvement.csv	Paper-ready improvement summary
fai_detailed_improvement.csv	Full improvement analysis
fai_experiment_results.csv	Complete raw experimental data
fai_improvement_summary.csv	Improvement over all baselines
fai_paper_results_table.csv	Ready for manuscript inclusion

---

🚀 Quick Start

☁️ Option 1: Google Colab (Recommended)

1. Click the badge below to open the notebook

2. Upload your adult.csv file when prompted
3. Run all cells — results and figures are generated automatically

💻 Option 2: Run Locally

# Clone the repository
git clone https://github.com/Junaid-Ahmed-Rupok/FAI-Feature-Wise-Adaptive-Imputation.git
cd FAI-Feature-Wise-Adaptive-Imputation

# Install dependencies
pip install -r requirements.txt

# Launch Jupyter
jupyter notebook

⚙️ Requirements

pip install -r requirements.txt

Package	Minimum Version
pandas	1.3.0
numpy	1.21.0
matplotlib	3.4.0
seaborn	0.11.0
scikit-learn	1.0.0

---

📖 Citation

If you use FAI in your research, please cite:

@article{emmanuel2021survey,
  title     = {A survey on missing data in machine learning},
  author    = {Emmanuel, Tlamelo and others},
  journal   = {Journal of Big Data},
  volume    = {8},
  number    = {140},
  year      = {2021},
  publisher = {Springer}
}

@misc{ahmed2024fai,
  author    = {Ahmed, Junaid},
  title     = {FAI: Feature-Wise Adaptive Imputation},
  year      = {2024},
  publisher = {GitHub},
  url       = {https://github.com/Junaid-Ahmed-Rupok/FAI-Feature-Wise-Adaptive-Imputation}
}

---

👨‍💻 Author

Sarder Junaid Ahmed

Data Scientist & Machine Learning Engineer

Transforming complex data into strategic decisions through rigorous statistical modeling and production-ready machine learning systems.

Specializations: Statistical ML · Causal Inference · Trustworthy AI · Fairness-Aware ML · RAG Systems

Selected Research:

• 📄 Ahmed, S.J. et al. (2026). Machine Learning for Crime Classification: A Fairness-Aware Approach to Class Imbalance. Journal of Machine Learning and Applications, 2(1), 9–17. DOI: 10.61577/jmla.2026.100002
• 📄 Ahmed, S.J. et al. (2026). Machine Learning for Crime Classification: A Fairness-Aware Approach to Class Imbalance. IEEE SPICSCON 2026, BAUET, Bangladesh (Aug 13–14, 2026). Accepted for Presentation — IEEE Xplore.
• 📄 Ahmed, S.J. et al. (2026). CF-EGAT: A Causal Fairness-Aware Equity Graph Attention Network for Country-Level Environmental Livability Classification. SPECTRA 2026. 🏆 1st Best Paper Award
• 📄 Ahmed, S.J. (2025). Multi-Dimensional Statistical Similarity for Governance Classification: Beyond Arbitrary Thresholds. APMEE 2025. 🏆 Best Research Paper Award
• 📄 Ahmed, S.J. (2026). DeepEnMap: Ordinal-Aware Multi-Modal Deep Learning for Energy Poverty Risk Mapping. IEMIS 2026, University of British Columbia, Vancouver, Canada (Aug 10–12, 2026). Accepted for Presentation — Springer LNNS Series (Scopus, EI-Compendex, DBLP, ISI Proceedings).
• 📄 Ahmed, S.J. (2026). Density-Decoupled, Mask-Ablated Segmentation-Guided Diffusion for Controllable Mammography Synthesis: A Preliminary Study. IEMIS 2026, University of British Columbia, Vancouver, Canada (Aug 10–12, 2026). Accepted for Presentation — Springer LNNS Series (Scopus, EI-Compendex, DBLP, ISI Proceedings).
• 📄 Ahmed, S.J., Islam Nahian, M.T., & Kwoshik, M.H.R. (2026). Environmental Livability Assessment via Adaptive Bootstrap-Retrained SHAP and Statistically-Constrained Pareto Counterfactuals: A Cross-National Analysis. IEEE SPICSCON 2026, BAUET, Bangladesh (Aug 13–14, 2026). Accepted for Presentation — IEEE Xplore.
• 📄 Ahmed, S.J. (2026). DemocracyGuard: Testing a Divergence-Index Reconciliation of Subjective and Objective Democracy Indicators for Forecasting Adverse Regime Transitions. Under Review, Transactions on Machine Learning Research (TMLR) — Q1, Top-Tier Journal.
• 📄 Ahmed, S.J. (2026). FAI: Feature-Wise Adaptive Imputation via Downstream-Aware Method Selection. Under Review, ICISET 2026 (IEEE Xplore).

Other Deployed Projects:

• 🔬 ReproHub — Automated research reproducibility platform with composite scoring across 11 statistical tests
• 📊 StatsPro — AI-powered statistical analysis platform with automated CSV-to-report workflows

Honors: 🏆 1st Best Paper — SPECTRA 2026  ·  🏆 Best Research Paper — APMEE 2025  ·  🎖️ Esteemed Alumni Award — YLRL RUET 2024  ·  ⭐ Perfect GPA 5.00/5.00 — SSC & HSC  ·  🎓 National Merit Scholarship — 2009 & 2013

---

📜 License

This project is licensed under the MIT License — see the LICENSE file for details.

---

⭐ Support This Project

If FAI helped your research, consider:

"The best imputation method is not the same for every feature — FAI learns to choose wisely."

Built with ❤️ by Junaid Ahmed

```