LightGBM in ML Python - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Understanding LightGBM's Leaf-wise Tree Growth

LightGBM grows trees using a leaf-wise strategy instead of a level-wise one. What is the main advantage of this leaf-wise growth compared to level-wise growth?

A It ensures all leaves grow equally, maintaining balanced trees.

B It reduces overfitting by limiting tree depth strictly.

C It uses random feature selection to reduce training time.

D It grows deeper trees with fewer splits, improving accuracy and efficiency.

💡 Hint

Think about how growing the leaf with the largest loss reduction affects tree shape and performance.

Leaf-wise growth chooses the leaf with the highest loss to split, allowing deeper trees where needed, which can improve accuracy and reduce training time compared to level-wise growth.

❓ Hyperparameter

intermediate

Choosing the Right LightGBM Hyperparameter for Overfitting Control

Which LightGBM hyperparameter primarily controls the maximum depth of the trees to prevent overfitting?

A max_depth

B num_leaves

C min_data_in_leaf

D learning_rate

💡 Hint

This parameter limits how deep the tree can grow.

max_depth sets the maximum depth of each tree, directly limiting complexity and helping prevent overfitting.

❓ Predict Output

advanced

Output of LightGBM Training with Early Stopping

What will be the output of the following Python code snippet using LightGBM?

ML Python

import lightgbm as lgb
from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.metrics import accuracy_score

X, y = load_breast_cancer(return_X_y=True)
X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, random_state=42)
train_data = lgb.Dataset(X_train, label=y_train)
val_data = lgb.Dataset(X_val, label=y_val, reference=train_data)

params = {'objective': 'binary', 'metric': 'binary_logloss', 'verbose': -1}

model = lgb.train(params, train_data, num_boost_round=100, valid_sets=[val_data], early_stopping_rounds=10, verbose_eval=False)
preds = model.predict(X_val, num_iteration=model.best_iteration)
preds_binary = (preds > 0.5).astype(int)
acc = accuracy_score(y_val, preds_binary)
print(f"Accuracy: {acc:.3f}")

A Accuracy: 0.965

B Accuracy: 0.850

C Accuracy: 0.500

D Raises a TypeError due to wrong parameter

💡 Hint

The breast cancer dataset is easy to classify with LightGBM, expect high accuracy.

The code trains LightGBM with early stopping on a well-known dataset, resulting in high accuracy around 0.96.

❓ Metrics

advanced

Interpreting LightGBM's Binary Logloss Metric

During LightGBM training for a binary classification task, the binary_logloss metric decreases from 0.6 to 0.2. What does this change indicate about the model?

A The model's predictions are becoming less accurate.

B The model's predicted probabilities are closer to true labels.

C The model is overfitting the training data.

D The model's accuracy is exactly 80%.

💡 Hint

Lower logloss means better probability estimates.

Binary logloss measures how close predicted probabilities are to actual labels; a decrease means better predictions.

🔧 Debug

expert

Identifying the Cause of a LightGBM Training Error

You run the following LightGBM training code but get the error: "ValueError: Label must be 1-D array." What is the most likely cause?

ML Python

import lightgbm as lgb
import numpy as np

X = np.random.rand(100, 5)
y = np.random.rand(100, 1)  # shape is (100,1)
train_data = lgb.Dataset(X, label=y)
params = {'objective': 'regression'}
model = lgb.train(params, train_data, num_boost_round=10)

A The num_boost_round value is too low to start training.

B The feature matrix X has too many columns for LightGBM.

C The label array y should be 1-dimensional, but it is 2-dimensional.

D The objective parameter 'regression' is invalid in LightGBM.

💡 Hint

Check the shape of the label array passed to Dataset.

LightGBM expects labels as a 1-D array. Passing a 2-D array causes this ValueError.