From prediction to interpretation
The manuscript analyzes the Kaggle NBA Games dataset, approximately 668,000 rows, using Game Score, regression models, SHAP, and clustering. The initial target was individual on-court PLUS_MINUS.
Data / AI & ML
NBA analytics: beyond prediction
Game Score, explainable machine learning, and six performance-based player clusters.
MY ROLEResearcher · Manuscript author
STATUSResearch project
YEAR2025
Research project
The result that changed the approach
The draft reports a best regression R² of approximately 0.14. Instead of presenting weak prediction as a success, the work pivots toward explaining model associations and identifying six performance-based clusters.
A careful reading of the findings
SHAP explains fitted-model behavior; it does not establish that a statistic causes a team to win or lose. The reported numbers are manuscript results, not an independently reproduced benchmark.
✳KEEP EXPLORING
A few connected ideas.
Research projectAI-powered exam preparation
Exploring how intelligent systems can support preparation for Mongolia’s General Entrance Examination.
Next.jsFastAPIPostgreSQL
Presented · Research projectAI-powered skin analysis
An application-oriented research project exploring AI-based skin image analysis.
MobileNetV2YOLOv8Computer Vision
Research project
IoT smart waste management
Exploring connected systems for smarter waste management.
IoTDataResearch
GOOD THINGS START WITH A HELLO
Have something
in mind?
A project, a research question, a talk, or a community idea — I’d love to hear about it.
PurevjargalResearcher & software engineer
Ulaanbaatar, Mongolia · UTC+8