This week, I worked on Evaluation Metrics, deepening my understanding of how to assess classification models beyond accuracy.
Using Python, Pandas, NumPy, Matplotlib, and Scikit-learn, I continued working with the 2026 Lead Scoring dataset and Logistic Regression to predict whether a lead converts.
Some of the key concepts I practiced include:
🔹 ROC AUC and evaluating numerical features
🔹 Precision and Recall across different classification thresholds
🔹 F1 score and threshold optimization
🔹 Visualizing precision and recall trade-offs
🔹 5-fold cross-validation
🔹 Hyperparameter tuning using different values of C
🔹 Evaluating model performance using mean ROC AUC and standard deviation
One key takeaway from this week was that evaluating a classification model goes beyond accuracy. Metrics such as ROC AUC, precision, recall, and F1 score provide different perspectives on model performance, while cross-validation helps assess how consistently a model performs across different data splits.
I also reinforced the importance of debugging and ensuring that notebook cells execute in the correct order to obtain consistent results.
I'm continuing to build consistency, strengthen my machine learning foundations, and apply what I learn through hands-on projects.
📂 I've documented my Week 4 learning and homework in my GitHub repository: https://github.com/giftyhealthtech/ml-zoomcamp-2026/tree/main/04-evaluation
On to the next week!
Top comments (0)