Wine KNN

In this data science project, I developed a classification model using the K-Nearest Neighbors (KNN) algorithm—entirely from scratch. This was a deep dive into the core mechanics of distance-based classification, focused on clarity, control, and precision at each step of the ML pipeline.

Highlights

• Algorithm Implementation: Crafted the full KNN algorithm manually without libraries to build conceptual understanding and flexibility.
• Data Analysis: Conducted EDA, cleaned the dataset, and engineered features for better decision boundaries.
• Nested Cross-Validation: Employed nested CV for robust model evaluation and hyperparameter tuning.
• Noise Resilience: Introduced controlled noise to test model robustness, simulating real-world scenarios with imperfect data.

Outcomes

This project delivered a fully custom KNN classifier with structured evaluation, optimized performance, and real-world robustness considerations. It served as a strong foundation for future classification projects and applied statistical learning.

Attachments