MilitaryForU
Personalized Military Specialty Recommendation
Overview
MilitaryForU helps prospective service members explore military specialties suited to their characteristics and skills.
Developed using public military specialty data and a custom-built tag dataset, the service combines collaborative filtering with Random Forest and XGBoost to analyze user similarity alongside specialty-specific eligibility requirements.
The project received the Top Excellence Award at the 2024 Defense Public Data Utilization Competition.
Problem
During my military service, I saw soldiers struggle in roles that did not match their strengths or interests. Such mismatches can make adaptation difficult and prevent the military from fully utilizing individual capabilities.
These observations motivated MilitaryForU: a service designed to help prospective service members explore suitable specialties by considering both personal fit and eligibility requirements.
My role
- Led project planning and team coordination.
- Built a custom tag dataset and developed the recommendation pipeline using public military specialty data.
- Implemented collaborative filtering to generate recommendations based on similar user profiles.
- Applied Random Forest and XGBoost in a subsequent analysis stage to address overfitting concerns and incorporate specialty-specific eligibility requirements.
Approach
- Questionnaire responses are mapped to tags representing users’ characteristics and skills.
- Collaborative filtering then generates initial recommendations by comparing each user with similar profiles.
- The initial approach encountered overfitting and did not adequately account for specialties requiring particular qualifications or academic majors.
- To address these limitations, Random Forest and XGBoost were applied in a subsequent analysis stage, using the collaborative filtering results alongside additional eligibility conditions to produce the final recommendations.
System architecture
- Data foundation: Public military specialty data and a custom-built tag dataset.
- User profiling: Questionnaire responses mapped to tags describing individual characteristics and skills.
- Initial recommendation: Collaborative filtering based on similar user profiles.
- Subsequent analysis: Random Forest and XGBoost analyze the initial recommendations alongside specialty-specific eligibility conditions.
- Output: Personalized military specialty recommendations.
Technologies
Results
- Received the Top Excellence Award at the 2024 Defense Public Data Utilization Competition.