👋 Hi, I’m Yejun
⸻
🧠 About Me
I focus on data science, machine learning, and applied mathematics, especially in sports analytics and computer vision. I prefer projects with clear metrics, testable assumptions, and interpretable results.
I primarily work in Python and care more about why a model works (or fails) than blindly improving accuracy.
⸻
🔎 What I’m Interested In (Click to Expand)
Machine Learning
• Feature engineering and leakage control
• Model evaluation and baselines
• Trade-offs between accuracy and interpretability
Sports Analytics
• Player performance modeling
• Metric design and validation
• Translating raw stats into decision-ready insights
Computer Vision
• Action classification from video
• Motion, speed, and occlusion challenges
• Dataset quality over model complexity
⸻
🚀 Featured Projects
⚽ Soccer Performance Prediction
Goal: Predict future player stats using historical match and season-level data.
Key Work • Rolling-window and lag-based features • Regression vs classification framing • Error analysis by player role and minutes played
Skills Used: Python, pandas, scikit-learn
⸻
⚾ MLB OBP-Weighted Offensive Value Model
Goal: Build a more realistic offensive contribution metric using OBP-weighted run creation ideas.
Key Work • Custom metric formulation • Comparison against traditional sabermetrics • Sensitivity analysis on weight choices
Skills Used: Data analysis, statistics, model validation
⸻
🏸 Badminton Action Classification
Goal: Classify badminton actions (smash, clear, drop, etc.) from video.
Key Work • Action labeling and dataset design • Frame-level vs sequence-level modeling • Performance trade-offs under motion blur
Skills Used: OpenCV, ML pipelines, video preprocessing
⸻
🛠 Tech Stack
Languages
Python
Libraries / Tools
NumPy, pandas, scikit-learn OpenCV Git & GitHub
Core Concepts
Feature engineering Model evaluation Metric design Bias & failure analysis
⸻
📊 GitHub Stats
⸻
📌 How I Work • Start with a baseline before complex models • Question the data before tuning hyperparameters • Prefer explainable failure over unexplained success
⸻
📬 Contact • 📧 School Email: @stu.siskorea.org
⸻
GitHub is where I test ideas, break assumptions, and refine models — not just where I store finished code.