Military Salary Estimator - Bubulka Analytics Portfolio

96%

Model Accuracy
R² = 0.9627

3,589

Training Records
Real transitions

±$3,246

Avg Error
MAE on test set

Live

Interactive
Shiny Dashboard

🚀 Quick Links

GitHub Repository

View source code, documentation, and contribute

Contributing

Development guidelines and how to contribute

📊 Project Overview

This project analyzes 3,589 real military-to-civilian salary transitions to build a predictive model that helps service members understand expected civilian salary outcomes. The model achieves 96% accuracy using a Generalized Linear Model (GLM) trained with proper cross-validation.

Military rank dominates salary determination, explaining 95%+ of variance. This reflects direct salary-based rank correspondence and the alignment between military hierarchy and civilian career progression.

✨ Key Features

📈 High Accuracy

R² = 0.9627 on independent test set. Achieves 96% variance explanation with minimal overfitting.

🔍 Interpretable

Transparent GLM coefficients explain exactly which factors drive salary outcomes.

🎯 Interactive

Live Shiny dashboard for real-time salary estimation with confidence intervals.

🔄 Reproducible

Full analysis pipeline with cross-validation, testing, and comprehensive documentation.

🔐 Privacy-Aware

No personally identifiable information. Military data anonymized and aggregated.

📝 Open Source

MIT licensed. Full source code available on GitHub for transparency and collaboration.

🛠️ Technical Stack

Algorithm: Generalized Linear Model (GLM)
Language: R 4.0+
Dashboard: Shiny interactive framework
Training Size: 2,512 records (70%)
Test Size: 1,077 records (30%)
Validation: 10-fold cross-validation
Explainability: SHAP & LIME analysis
Visualization: ggplot2 publication-quality figures

🔬 Model Insights

Feature Importance

The model reveals clear feature importance rankings:

Military Rank: 95.2% - Dominates salary determination
Years of Service: 3.1% - Secondary factor
Occupation: 1.4% - Minor contribution
Civilian Category: 0.2% - Minimal impact
Skill Level: 0.1% - Negligible contribution

Key Findings

Rank directly encodes compensation bands—higher ranks earn significantly more
Years of service has modest secondary effect on salary
Model is near-optimal; additional complexity provides minimal improvement
Skills feature explored but found redundant with occupation coding

💡 Use Cases

Military-to-civilian career transition planning
Salary expectation setting for separating service members
Career counselor reference tool for transitioning veterans
Veterans employment program benchmarking
Research on military-civilian workforce transitions

📋 Data & Methodology

Data Source

3,589 real military-to-civilian transitions with complete records. Data quality: 0 duplicates, <1% missing.

Train/Test Split

Training: 2,512 records (70%)
Testing: 1,077 records (30% holdout)
Method: Stratified random split

Model Validation

Cross-Validation: 10-fold with R² = 0.8202 ± 0.0304 (conservative)
Test Set: R² = 0.9627 (96.27% variance explained)
Overfitting Check: 0.02% train→test drop (minimal)
Residual Analysis: Normally distributed, homoscedastic

Performance Metrics

RMSE: $5,003
MAE: $3,246
Prediction Range: $28,000 - $180,000
Mean Salary: $68,450

🚀 Getting Started

Clone the repository and get the model running in minutes:

1. Clone Repository


                git clone https://github.com/yourusername/military-salary-estimator.git

2. Install Dependencies


                # In R console:

                source("requirements.R")

3. Launch Dashboard


                library(shiny)

                runApp("10_shiny_dashboard/app_simple.R", port = 8100)

Visit http://127.0.0.1:8100 in your browser. Done! 🎉

⚠️ Limitations & Disclaimer

This model is a powerful tool for understanding salary trends, but comes with important limitations:

Selection Bias: Only includes successful transitions; may overestimate for struggling separatees
Cross-Sectional: Snapshot in time; doesn't capture career progression
Missing Variables: Education, location/COL, benefits not included
Occupational Matching: ±3-5% error in niche specialties
Regional Variation: Ignores geographic salary differences (future enhancement)

Use responsibly: Combine with BLS data, career counseling, and personal judgment.

📜 License & Attribution

This project is licensed under the MIT License, allowing free commercial and research use with attribution.

Citation:


                @software{military_salary_estimator,

                  title = {Military-to-Civilian Salary Estimator},

                  author = {Your Name},

                  year = {2024},

                  url = {https://github.com/yourusername/military-salary-estimator}

                }

Ready to explore?

Check out the repository on GitHub or launch the interactive dashboard to see the model in action.

View on GitHub Get Started

🎯 Military-to-Civilian Salary Estimator

96%