CatBoost in ML Python - Cheat Sheet & Quick Revision

Recall & Review

beginner

What is CatBoost?

CatBoost is a machine learning algorithm that builds decision trees and is especially good at handling categorical data without needing to convert it manually.

intermediate

How does CatBoost handle categorical features?

CatBoost automatically converts categorical features into numbers using a special technique called 'ordered target statistics' which helps avoid overfitting.

intermediate

What is the main advantage of using CatBoost over other gradient boosting methods?

CatBoost reduces prediction shift and overfitting by using ordered boosting and supports categorical data natively, making it faster and more accurate on many datasets.

beginner

What metric can you use to evaluate a CatBoost model for classification?

You can use accuracy, AUC (Area Under the Curve), or log-loss to evaluate a CatBoost classification model depending on your problem.

advanced

Explain the concept of 'ordered boosting' in CatBoost.

Ordered boosting is a technique where CatBoost builds trees using a special permutation of data to prevent target leakage and reduce overfitting during training.

What type of data does CatBoost handle natively without manual preprocessing?

A Only numerical data

B Categorical data

C Text data only

D Image data

CatBoost is designed to handle categorical data directly, unlike many other algorithms that require manual encoding.

Which technique does CatBoost use to avoid overfitting when processing categorical features?

A Ordered target statistics

B Label encoding

C One-hot encoding

D Random sampling

CatBoost uses ordered target statistics to convert categorical features into numbers while reducing overfitting.

What is the main purpose of ordered boosting in CatBoost?

A Prevent target leakage and reduce overfitting

B Speed up training

C Increase model size

D Simplify data preprocessing

Ordered boosting helps prevent target leakage by using permutations of data during training, which reduces overfitting.

Which of the following is NOT a typical evaluation metric for CatBoost classification models?

A Log-loss

B Accuracy

C AUC

D Mean Squared Error

Mean Squared Error is mainly used for regression, not classification.

CatBoost is a type of which machine learning method?

A Neural network

B K-nearest neighbors

C Gradient boosting

D Support vector machine

CatBoost is a gradient boosting algorithm that builds decision trees sequentially.

Describe how CatBoost handles categorical features and why this is beneficial.

Think about how CatBoost avoids manual steps and overfitting with categorical data.

Explain the concept of ordered boosting and its role in CatBoost's training process.

Consider how training order affects model learning and overfitting.