Data science Screen Test

By mfh.officials@gmail.com / January 7, 2025

Data science

Data science Screening test

1 / 20

Which of the following is TRUE about ensemble methods?

They cannot be used with decision trees

They combine multiple weak models to create a stronger model

They are less prone to overfitting than single models

They are always more computationally expensive than single models

2 / 20

What is the purpose of the Adam optimizer in neural networks?

To perform backpropagation

To adjust the learning rate during training

To calculate the gradient of the loss function

To reduce the loss function

3 / 20

Which of the following techniques is used for dimensionality reduction?

Principal Component Analysis (PCA)

Decision Trees

K-means clustering

Naive Bayes

4 / 20

Which of the following is an example of unsupervised learning?

Principal Component Analysis (PCA)

Linear Regression

Logistic Regressi

Decision Trees

5 / 20

2. In a decision tree, the split criterion is typically based on:

Variance

Sum of squared errors

Root mean squared error

Information gain or Gini impurity

6 / 20

Which of the following is the most appropriate way to deal with missing values in a dataset?

Replace missing values with the mean or median of the column

Drop all rows with missing values

Replace missing values with zeros

Use a model-based imputation method

7 / 20

Which of the following libraries is primarily used for deep learning?

TensorFlow

scikit-learn

Matplotlib

pandas

8 / 20

What is the "bias-variance tradeoff"?

Increasing model complexity reduces both bias and variance

Increasing model complexity reduces bias and increases variance

Increasing model complexity reduces variance and increases bias

Increasing model complexity does not affect bias or variance

9 / 20

Which of the following is a disadvantage of the k-nearest neighbors (KNN) algorithm?

Answer: B) It requires a large amount of training data

It cannot handle multi-class classification

It is difficult to interpret

It assumes linearity in the data

It requires a large amount of training data

10 / 20

What is the purpose of regularization in machine learning?

To speed up training

To increase model complexity

To improve model interpretability

To reduce overfitting by penalizing large coefficients

11 / 20

Which of the following metrics is used to evaluate clustering models?

Silhouette score

Precision

F1-score

ROC-AUC

12 / 20

In the context of model evaluation, what does the "ROC curve" stand for?

Root Output Curve

Recurrent Operations Curve

Residual Output Curve

Receiver Operating Characteristic Curve

13 / 20

Which evaluation metric is most appropriate for imbalanced classification problems?

R-squared

Precision and Recall

Accuracy

Mean Squared Error

14 / 20

Which of the following is a hyperparameter for the k-means clustering algorithm?

Activation function

Learning rate

Number of clusters (k)

Regularization strength

15 / 20

In a time series forecasting problem, which of the following is most commonly used to check for stationarity?

Augmented Dickey-Fuller (ADF) test

Shapiro-Wilk test

ACF/PACF plots

Durbin-Watson test

16 / 20

Which of the following is a key assumption of the linear regression model?

Independence of dependent variables

Multicollinearity

Homoscedasticity

Non-linearity between independent and dependent variables

17 / 20

Which of the following algorithms is most appropriate for predicting a continuous outcome variable?

Linear regression

K-means clustering

K-nearest neighbors (classification)

Decision tree classification

18 / 20

The "curse of dimensionality" refers to:

The increasing complexity of models with more features

The difficulty of visualizing data in high-dimensional spaces

The difficulty in finding a suitable machine learning model

The tendency of high-dimensional data to become sparse

19 / 20

In random forests, what is the primary advantage over a single decision tree?

It is more interpretable

It always uses shallow trees

It uses more training data

It reduces variance and improves accuracy by averaging predictions

20 / 20

Cross-validation is primarily used to:

Reduce overfitting and assess model performance

Split data into training and testing sets

Tune hyperparameters

Visualize data

Your score is