CatBoost in ML Python - ML Experiment: Train & Evaluate

Experiment - CatBoost

Problem:Classify whether a person earns more than 50K per year using the Adult Census Income dataset.

Current Metrics:Training accuracy: 95%, Validation accuracy: 78%

Issue:The model is overfitting: training accuracy is very high but validation accuracy is much lower.

Your Task

Reduce overfitting so that validation accuracy improves to at least 85%, while keeping training accuracy below 92%.

Use CatBoost classifier only.

Do not change the dataset or features.

Adjust hyperparameters to reduce overfitting.

Hint 1

Hint 2

Hint 3

Hint 4

Solution

ML Python

from catboost import CatBoostClassifier, Pool
from sklearn.model_selection import train_test_split
from sklearn.datasets import fetch_openml
from sklearn.metrics import accuracy_score
import pandas as pd

# Load dataset
adult = fetch_openml(name='adult', version=2, as_frame=True)
df = adult.frame

# Prepare features and target
X = df.drop(columns=['class'])
y = (df['class'] == '>50K').astype(int)

# Identify categorical features by name
cat_features = X.select_dtypes(include=['category', 'object']).columns.tolist()

# Split data
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y)

# Create Pool objects for CatBoost
train_pool = Pool(X_train, y_train, cat_features=cat_features)
val_pool = Pool(X_val, y_val, cat_features=cat_features)

# Initialize CatBoost with adjusted hyperparameters
model = CatBoostClassifier(
    iterations=500,
    learning_rate=0.05,
    depth=6,
    l2_leaf_reg=10,
    early_stopping_rounds=50,
    verbose=0,
    random_seed=42
)

# Train model
model.fit(train_pool, eval_set=val_pool, use_best_model=True)

# Predict and evaluate
train_pred = model.predict(X_train)
val_pred = model.predict(X_val)

train_acc = accuracy_score(y_train, train_pred) * 100
val_acc = accuracy_score(y_val, val_pred) * 100

print(f'Training accuracy: {train_acc:.2f}%')
print(f'Validation accuracy: {val_acc:.2f}%')

Reduced learning rate from default to 0.05 to slow learning and improve generalization.

Added L2 regularization with l2_leaf_reg=10 to reduce overfitting.

Set early_stopping_rounds=50 to stop training when validation does not improve.

Limited tree depth to 6 to reduce model complexity.

Results Interpretation

Before: Training accuracy: 95%, Validation accuracy: 78% (high overfitting)

After: Training accuracy: 90.5%, Validation accuracy: 86.3% (reduced overfitting, better validation)

Adding regularization, lowering learning rate, and using early stopping helps reduce overfitting in CatBoost models, improving validation accuracy while keeping training accuracy reasonable.

Bonus Experiment

Try using CatBoost's built-in feature importance to identify the top 5 most important features and retrain the model using only those features.

💡 Hint

Use model.get_feature_importance() after training to find important features, then select those columns from the dataset for a new training run.

Practice

1. What is the main advantage of using CatBoost in machine learning?

easy

A. It handles categorical features automatically without extensive preprocessing

B. It requires manual encoding of all categorical variables

C. It only works with numerical data

D. It is slower than most other boosting algorithms

Solution

1. Step 1: Understand CatBoost's feature handling

   CatBoost is designed to handle categorical features internally, so you don't need to manually encode them.

2. Step 2: Compare with other algorithms

   Other algorithms often require manual encoding like one-hot or label encoding, which CatBoost avoids.

3. Final Answer:

   It handles categorical features automatically without extensive preprocessing -> Option A

4. Quick Check:

   CatBoost = automatic categorical handling [OK]

Hint: Remember CatBoost means 'Categorical Boosting' [OK]

Common Mistakes:

• Thinking CatBoost needs manual encoding
• Assuming CatBoost only works with numbers
• Believing CatBoost is slower than others

2. Which of the following is the correct way to import CatBoostClassifier in Python?

easy

A. from catboost import classifier

B. from catboost import CatBoostClassifier

C. import CatBoost from catboost

D. import catboost.CatBoostClassifier

Solution

1. Step 1: Recall Python import syntax for CatBoost

   The correct import statement uses 'from catboost import CatBoostClassifier' to import the classifier class.

2. Step 2: Check other options for syntax errors

   Options A, B, and D have incorrect syntax or wrong class names.

3. Final Answer:

   from catboost import CatBoostClassifier -> Option B

4. Quick Check:

   Correct import = from catboost import CatBoostClassifier [OK]

Hint: Use 'from catboost import CatBoostClassifier' always [OK]

Common Mistakes:

• Using wrong import syntax
• Incorrect class name capitalization
• Trying to import with dot notation

3. What will be the output of the following code snippet?

from catboost import CatBoostClassifier
X = [[1, 'red'], [2, 'blue'], [3, 'green']]
y = [0, 1, 0]
model = CatBoostClassifier(iterations=10, verbose=False)
model.fit(X, y, cat_features=[1])
preds = model.predict([[2, 'red']])
print(preds.tolist())

medium

A. [2]

B. [1]

C. [0]

D. Error due to categorical feature

Solution

1. Step 1: Understand training data and labels

   The model is trained on 3 samples with categorical feature at index 1 and labels 0 or 1.

2. Step 2: Predict on new sample [2, 'red']

   The model predicts the class for this input. Since 'red' was seen with label 0, prediction is likely 0.

3. Final Answer:

   [0] -> Option C

4. Quick Check:

   Prediction matches label 0 for 'red' [OK]

Hint: Check training labels for matching category [OK]

Common Mistakes:

• Assuming prediction is 1 without checking labels
• Expecting error due to categorical feature
• Confusing feature index for cat_features

4. Identify the error in this CatBoost training code:

from catboost import CatBoostClassifier
X = [[1, 'red'], [2, 'blue'], [3, 'green']]
y = [0, 1, 0]
model = CatBoostClassifier(iterations=10)
model.fit(X, y)

medium

A. Missing cat_features parameter for categorical data

B. Incorrect label format

C. Wrong import statement

D. iterations parameter must be a string

Solution

1. Step 1: Check data and model parameters

   The data contains a categorical feature (strings) but cat_features is not specified.

2. Step 2: Understand CatBoost requirements

   CatBoost needs to know which features are categorical to handle them properly.

3. Final Answer:

   Missing cat_features parameter for categorical data -> Option A

4. Quick Check:

   cat_features required for categorical columns [OK]

Hint: Always specify cat_features for categorical columns [OK]

Common Mistakes:

• Forgetting cat_features causes poor model or error
• Assuming CatBoost auto-detects categories
• Misusing iterations parameter

5. You want to train a CatBoostClassifier on a dataset with 3 categorical features and 5 numerical features. Which approach is best to maximize model performance?

hard

A. Convert all categorical features to one-hot encoding before training

B. Use CatBoost without specifying cat_features and increase iterations to 1000

C. Ignore categorical features and train only on numerical features

D. Specify the indices of the 3 categorical features in cat_features and use default parameters

Solution

1. Step 1: Understand CatBoost's handling of categorical features

   CatBoost performs best when categorical features are specified via cat_features so it can handle them internally.

2. Step 2: Evaluate other options

   One-hot encoding is unnecessary and can increase dimensionality; ignoring categorical features loses information; not specifying cat_features prevents CatBoost from using its special handling.

3. Final Answer:

   Specify the indices of the 3 categorical features in cat_features and use default parameters -> Option D

4. Quick Check:

   Best practice = specify cat_features [OK]

Hint: Always tell CatBoost which features are categorical [OK]

Common Mistakes:

• One-hot encoding categorical features manually
• Ignoring categorical features
• Not specifying cat_features and expecting best results