Class Central is learner-supported. When you buy through links on our site, we may earn an affiliate commission.

Fixing Classical Models – Diagnosis & Regularization

Go to class Write review

Overview

In this course, learners will improve a poorly performing classical ML model using core diagnostic and regularization techniques. The model starts off weak, and learners fix it step by step through evaluation, regularization, capacity tuning, and early stopping. All models are built using scikit-learn or XGBoost.

Syllabus

Unit 1: Evaluating Classification Models: Confusion Matrix and Classification Report

Fixing Models with One Parameter Change
Building a Confusion Matrix from Scratch
Calculating Classification Metrics by Hand
Fixing Misaligned Evaluation Metrics

Unit 2: Tuning L2 Regularization in Logistic Regression

Testing Regularization Strength in Action
Comparing L1 and L2 Regularization Effects
Refactoring Regularization for Better Reuse
Visualizing Regularization Effects with Matplotlib

Unit 3: Tuning Tree Depth to Prevent Overfitting in Decision Trees

Limiting Tree Depth to Prevent Overfitting
Tracking Tree Depth Performance Patterns
Fixing Swapped Metrics in Performance Reports
Automating Optimal Tree Depth Selection

Unit 4: Early Stopping with XGBoost: Preventing Overfitting in Boosted Trees

Implementing Early Stopping in XGBoost
Fixing Early Stopping Validation Bug
Using the Best Iteration for Predictions
Finding the Best Early Stopping Setting

Reviews

Start your review of Fixing Classical Models – Diagnosis & Regularization