Machine Learning

Semiconductor Wafer Defect Classification Using Machine Learning

The Semiconductor Wafer Defect Classification Using Machine Learning system is a comprehensive machine learning solution that classifies semiconductor wafers as defective or non-defective using process parameters including temperature, pressure, gas flow, and etch rate. This project is ideal for B.Tech, MCA, CSE, and engineering students as their final year project. It uses advanced regularized regression algorithms like Ridge and Lasso Regression with SMOTE to achieve 87.34% accuracy.

The system leverages 15 features including temperature, pressure, gas flow, etch rate, voltage, current, and process step information. Engineered features like temperature-pressure ratio, gas-etch ratio, and power enhance predictive capability. It provides interactive visualizations, feature importance analysis, model comparison, and real-time wafer defect classification to help semiconductor manufacturers improve yield and reduce scrap.

• Binary Wafer Defect Classification
• Ridge Regression (87.34% Accuracy)
• Lasso Regression (86.21% Accuracy)
• SMOTE for Class Imbalance
• Leak-Free Cross-Validation
• Feature Importance Analysis
• Real-time Defect Prediction
• Model Performance Comparison
• 3-Fold Stratified CV
• Flask Web Application