Machine Learning MCQ PDF

MCQ – Machine Learning – All Units 1. What does the True Positive (TP) value in a confusion matrix represent? a) Incorrectly predicted positive cases b) Correctly predicted negative cases c) Correctly predicted positive cases d) Incorrectly predicted negative cases Answer...

MCQ – Machine Learning – All Units 1. What does the True Positive (TP) value in a confusion matrix represent? a) Incorrectly predicted positive cases b) Correctly predicted negative cases c) Correctly predicted positive cases d) Incorrectly predicted negative cases Answer: c) Correctly predicted positive cases 2. Which metric is calculated using the formula Precision=TPTP+FP\text{Precision} = \frac{TP}{TP + FP}Precision=TP+FPTP? a) Recall b) Specificity c) Accuracy d) Precision Answer: d) Precision 3. If a model has high recall but low precision, what does it indicate? a) The model predicts all positives correctly but also includes many false positives. b) The model predicts all negatives correctly but misses some positives. c) The model predicts all negatives correctly but has many false negatives. d) The model is highly accurate overall. Answer: a) The model predicts all positives correctly but also includes many false positives. 4. What is the formula for the F1-score? a) 2×(Precision×Recall)Precision+Recall\frac{2 \times (\text{Precision} \times \text{Recall})}{\text{Precision} + \text{Recall}}Precision+Recall2×(Precision×Recall) b) Precision+Recall2\frac{\text{Precision} + \text{Recall}}{2}2Precision+Recall c) Precision×Recall\text{Precision} \times \text{Recall}Precision×Recall d) TPTP+FN\frac{TP}{TP + FN}TP+FNTP Answer: a) 2×(Precision×Recall)Precision+Recall\frac{2 \times (\text{Precision} \times \text{Recall})}{\text{Precision} + \text{Recall}}Precision+Recall2×(Precision×Recall)  5. If a confusion matrix shows many False Negatives, what does it imply about the model? a) It is overly cautious in predicting negatives. b) It is missing many actual positive cases. c) It is overfitting to the training data. d) It is achieving high precision. Answer: b) It is missing many actual positive cases. 6. Which metric is most appropriate to evaluate a model for an imbalanced dataset? a) Accuracy b) F1-score c) Mean Absolute Error d) R-squared Answer: b) F1-score 7. What does the Specificity metric measure in a confusion matrix? a) The proportion of true negatives correctly identified. b) The proportion of true positives correctly identified. c) The proportion of all predictions that are correct. d) The proportion of false positives misclassified. Answer: a) The proportion of true negatives correctly identified. 8. Which of the following statements is true about Accuracy in a confusion matrix? a) Accuracy is always the best metric for evaluating a classifier. b) Accuracy is calculated as TPTP+FN\frac{TP}{TP + FN}TP+FNTP. c) Accuracy can be misleading for imbalanced datasets. d) Accuracy is equal to Precision for all classifiers. Answer: c) Accuracy can be misleading for imbalanced datasets. 9. In a fraud detection system, which metric would be more important to minimize the cost of false positives and false negatives? a) Recall b) Precision c) F1-score d) Specificity Answer: c) F1-score  10. If a model predicts all cases as positive, which metrics will be affected? a) Only Precision and Recall b) Precision, Recall, and Specificity c) Only Recall d) Only Specificity Answer: b) Precision, Recall, and Specificity Scenario: Loan Approval Prediction You are working on a dataset with the following attributes: Income: Annual income of the applicant. Loan Amount: Amount requested for the loan. Credit History: 1 if the applicant has a good credit history, 0 otherwise. Loan Approved: 1 if the loan is approved, 0 otherwise. You aim to build a model using Orange to predict whether a loan will be approved. Q1: What is the first step after loading the dataset into Orange? a) Build a decision tree model immediately. b) Split the dataset into training and test sets. c) Normalize numerical features like Income and Loan Amount. d) Visualize data distributions and check for missing values. Answer: d) Visualize data distributions and check for missing values. Q2: If the dataset has missing values, what Orange widget would you use to handle them? a) Feature Constructor b) Data Sampler c) Impute d) Scatter Plot Answer: c) Impute Q3: After splitting the dataset, which metric is most suitable for evaluating model performance if the dataset is imbalanced? a) Accuracy b) Precision c) Recall d) F1-Score Answer: d) F1-Score Q4: You built a decision tree and logistic regression model in Orange. The decision tree gives an accuracy of 90%, while logistic regression gives 85%. How should you decide which model to deploy? a) Always choose the model with higher accuracy. b) Check metrics like precision, recall, and F1-score for both models before deciding. c) Use the simpler model, logistic regression, for interpretability. d) Randomly select one model. Answer: b) Check metrics like precision, recall, and F1-score for both models before deciding. Q5: Which Orange widget would you use to visualize the feature importance of the decision tree model? a) Box Plot b) Rank c) Scatter Plot d) Confusion Matrix Answer: b) Rank Scenario 1: Customer Churn Prediction You are working on a telecom dataset to predict whether a customer will churn (leave the company). The dataset contains features such as tenure, monthly charges, contract type, and churn (target: Yes/No). Q1: Which machine learning algorithm is best suited to predict whether a customer will churn? a) K-Means Clustering b) Logistic Regression c) Principal Component Analysis (PCA) d) Apriori Answer: b) Logistic Regression Explanation: Logistic regression is a classification algorithm that works well for binary outcomes like churn prediction. Q2: If the dataset has imbalanced classes with only 20% of customers labeled as "Yes" for churn, which metric should be prioritized? a) Accuracy b) Precision c) Recall d) Mean Absolute Error Answer: c) Recall Explanation: In churn prediction, missing true positives (actual churners) can be costly. Prioritizing recall ensures that most churners are correctly identified. Q3: To identify which features contribute most to churn, which technique would you use? a) Confusion Matrix b) Feature Importance (using Random Forest or similar algorithms) c) Cross-Validation d) k-Nearest Neighbors Answer: b) Feature Importance (using Random Forest or similar algorithms) Explanation: Random Forest provides a ranking of feature importance, helping to identify the most significant predictors of churn. Scenario 2: Fraud Detection in Financial Transactions You are tasked with building a model to detect fraudulent transactions. The dataset includes features like transaction amount, time, geographic location, and fraud (target: Yes/No). Q4: Which algorithm is most suitable for detecting fraudulent transactions in real-time? a) Naive Bayes b) Gradient Boosting (e.g., XGBoost) c) K-Means Clustering d) Linear Regression Answer: b) Gradient Boosting (e.g., XGBoost) Explanation: Gradient boosting algorithms are powerful for tabular data and can efficiently handle imbalanced datasets often found in fraud detection. Q5: The dataset has 95% non-fraudulent transactions and 5% fraudulent ones. What approach can you use to address this imbalance? a) Remove non-fraudulent transactions to balance the dataset. b) Use accuracy as the evaluation metric. c) Apply oversampling (e.g., SMOTE) or undersampling techniques. d) Normalize the dataset to reduce class imbalance. Answer: c) Apply oversampling (e.g., SMOTE) or undersampling techniques. Explanation: Oversampling or undersampling can balance the dataset and help algorithms learn minority class patterns effectively. Q6: After training your model, you observe a high false positive rate. What is the best way to address this? a) Increase the dataset size. b) Use a different algorithm, such as k-Nearest Neighbors. c) Adjust the classification threshold to reduce false positives. d) Ignore the false positive rate and focus on accuracy. Answer: c) Adjust the classification threshold to reduce false positives. Explanation: Tweaking the threshold can shift the balance between precision and recall, potentially reducing false positives. Scenario 3: Recommender System for E-Commerce An e-commerce platform wants to recommend products to its users based on their browsing and purchase history. Q7: Which algorithm is most suitable for this scenario? a) Support Vector Machines b) Collaborative Filtering c) Decision Trees d) Naive Bayes Answer: b) Collaborative Filtering Explanation: Collaborative filtering is commonly used for recommendation systems, leveraging user-item interactions to provide personalized recommendations. Q8: To handle the cold-start problem for new users who have no interaction history, which method should be used? a) Content-Based Filtering b) Collaborative Filtering c) K-Means Clustering d) Random Forest Answer: a) Content-Based Filtering Explanation: Content-based filtering uses the features of items (e.g., category, price) to recommend products, bypassing the need for user interaction history. Scenario 4: Image Classification for Disease Detection You are tasked with building a model to classify chest X-ray images as normal or diseased. Q9: Which algorithm or architecture is most suitable for this task? a) Logistic Regression b) Convolutional Neural Network (CNN) c) Decision Trees d) k-Nearest Neighbors Answer: b) Convolutional Neural Network (CNN) Explanation: CNNs are designed for image data and are highly effective for tasks like image classification. Q10: During training, the model achieves high accuracy on training data but performs poorly on validation data. What is the most likely issue? a) Overfitting b) Underfitting c) Lack of data normalization d) Incorrect feature selection Answer: a) Overfitting Explanation: High training accuracy and poor validation performance indicate overfitting, where the model learns patterns specific to the training set but fails to generalize. Scenario 1: Disease Diagnosis (Classification) A hospital uses a dataset to predict whether a patient has diabetes (1) or not (0). The dataset includes features like age, BMI, blood pressure, and glucose levels. Q1: Which machine learning algorithm is most appropriate for this task? a) Linear Regression b) Logistic Regression c) K-Means Clustering d) PCA Answer: b) Logistic Regression Explanation: Logistic regression is used for binary classification problems like predicting the presence or absence of a disease. Q2: If the dataset contains 80% "No Diabetes" and 20% "Diabetes," which metric is most important to evaluate the model? a) Accuracy b) Precision and Recall c) Mean Absolute Error d) R-squared Answer: b) Precision and Recall Explanation: For imbalanced datasets, accuracy can be misleading. Precision and recall help assess the model’s performance on the minority class (diabetes). Q3: During training, your classifier achieves 95% accuracy, but you find that it predicts "No Diabetes" for almost all cases. What might be the issue? a) Overfitting b) Data imbalance c) Underfitting d) Model complexity Answer: b) Data imbalance Explanation: The model is biased toward the majority class due to the imbalance in the dataset. Q4: To improve the model's ability to detect diabetes, what approach would you take? a) Use SMOTE (Synthetic Minority Oversampling Technique) b) Increase the size of the dataset by adding more "No Diabetes" cases c) Apply PCA for dimensionality reduction d) Normalize the dataset Answer: a) Use SMOTE (Synthetic Minority Oversampling Technique) Explanation: SMOTE generates synthetic samples for the minority class to balance the dataset. Scenario 2: Predicting House Prices (Regression) A real estate company wants to predict house prices based on features like area, number of bedrooms, location, and year built. Q5: Which algorithm would you use for this task? a) Decision Tree b) Linear Regression c) Logistic Regression d) K-Means Clustering Answer: b) Linear Regression Explanation: Linear regression is used for predicting continuous numerical values like house prices. Q6: If the relationship between features and house prices is nonlinear, which model might perform better than linear regression? a) k-Nearest Neighbors b) Random Forest Regressor c) Gradient Boosting Regressor d) All of the above Answer: d) All of the above Explanation: Nonlinear models like k-NN, Random Forest, and Gradient Boosting can capture complex relationships between features and target variables. Q7: The R-squared score of your model is 0.95 on training data and 0.60 on test data. What does this indicate? a) Underfitting b) Overfitting c) Good generalization d) High bias Answer: b) Overfitting Explanation: The model performs well on training data but poorly on test data, indicating it is overfitting to the training set. Q8: If the dataset has a wide range of feature scales (e.g., area in square meters vs. number of bedrooms), which preprocessing step is necessary? a) One-hot encoding b) Normalization or Standardization c) Data augmentation d) Feature selection Answer: b) Normalization or Standardization Explanation: Scaling ensures that features with different ranges do not disproportionately influence the model. Scenario 3: Customer Satisfaction Prediction (Regression and Classification) A company wants to predict customer satisfaction scores on a scale from 1 to 10 and classify customers as satisfied (scores 7–10) or unsatisfied (scores 1–6). Q9: Predicting customer satisfaction scores (1–10) is an example of: a) Regression b) Classification c) Clustering d) Dimensionality Reduction Answer: a) Regression Explanation: Predicting numerical scores (1–10) involves regression. Q10: Classifying customers as satisfied or unsatisfied is an example of: a) Regression b) Binary Classification c) Multi-Class Classification d) Clustering Answer: b) Binary Classification Explanation: Dividing customers into two categories (satisfied or unsatisfied) is a binary classification task. Q11: You notice that customer satisfaction prediction errors are higher for scores at the extremes (1 and 10). Which metric should you use to evaluate the regression model? a) R-squared b) Mean Absolute Error (MAE) c) Root Mean Squared Error (RMSE) d) Accuracy Answer: c) Root Mean Squared Error (RMSE) Explanation: RMSE penalizes large errors more heavily than MAE, making it suitable for situations where extreme errors matter. Q12: For classification, which technique can help you identify whether satisfied customers are being correctly classified? a) Confusion Matrix b) PCA c) Feature Scaling d) K-Fold Cross-Validation Answer: a) Confusion Matrix Explanation: The confusion matrix provides detailed information about true positives, true negatives, false positives, and false negatives. Scenario 1: Employee Attrition Prediction A company wants to predict whether an employee will leave the organization (attrition) based on factors like job satisfaction, years at the company, salary level, and work-life balance. Q1: Why might a decision tree be a good choice for this task? a) It is robust to overfitting. b) It works well with large datasets only. c) It is interpretable and handles both numerical and categorical data. d) It requires data to be normalized. Answer: c) It is interpretable and handles both numerical and categorical data. Explanation: Decision trees are easy to interpret and can handle mixed data types without preprocessing like normalization. Q2: After training a decision tree, you notice that the model performs perfectly on training data but poorly on test data. What is the likely problem? a) The tree is too shallow. b) The tree is overfitting. c) The features are not normalized. d) There is too little data. Answer: b) The tree is overfitting. Explanation: A decision tree that performs perfectly on training data but poorly on unseen data is overfitting, likely because it is too deep and has memorized the training data. Q3: Which hyperparameter should you adjust to reduce overfitting in a decision tree? a) Maximum depth b) Learning rate c) Number of features d) Number of samples Answer: a) Maximum depth Explanation: Limiting the maximum depth of the tree prevents it from growing too complex, which helps avoid overfitting. Q4: If you want to find out which feature is the most important in predicting attrition, what should you look at? a) Gini Impurity or Information Gain of each split b) Accuracy of the model c) Leaf nodes of the tree d) Confusion matrix Answer: a) Gini Impurity or Information Gain of each split Explanation: Decision trees use metrics like Gini Impurity or Information Gain to determine which features contribute the most to the splits, indicating their importance. Scenario 2: Loan Default Prediction A bank uses a dataset to predict whether a loan applicant will default on their loan. Features include income, loan amount, credit score, and employment status. Q5: Which splitting criterion would you use in a decision tree to predict loan defaults? a) Mean Squared Error b) Gini Index or Entropy c) R-squared d) Cosine Similarity Answer: b) Gini Index or Entropy Explanation: Gini Index and Entropy are splitting criteria used in decision trees for classification tasks like predicting loan defaults. Q6: Your decision tree model achieves 85% accuracy, but the bank is more concerned about minimizing the number of loan defaults (false negatives). What should you focus on instead of accuracy? a) Precision b) Recall c) F1-Score d) R-squared Answer: b) Recall Explanation: Recall ensures that most loan defaulters are correctly identified, reducing false negatives. Q7: When training the decision tree, you observe the following: The model performs poorly on both training and test data. What is the likely issue? a) The tree is too deep. b) The tree is too shallow. c) The dataset is imbalanced. d) The features are not scaled. Answer: b) The tree is too shallow. Explanation: A shallow tree cannot capture the complexity of the data, leading to underfitting and poor performance on both training and test data. Q8: If a decision tree is performing well but you want to make it more robust and less sensitive to data variations, what technique should you use? a) Gradient Boosting b) Bagging (e.g., Random Forest) c) Normalization d) Feature Scaling Answer: b) Bagging (e.g., Random Forest) Explanation: Bagging combines multiple decision trees to create an ensemble model like Random Forest, which is more robust and less prone to overfitting. Scenario 3: Sales Prediction A retail company wants to predict the revenue for its stores based on features like store size, location, customer footfall, and average spend per customer. Q9: If the target variable is continuous (revenue), what splitting criterion would a decision tree use? a) Gini Impurity b) Entropy c) Mean Squared Error or Mean Absolute Error d) Information Gain Answer: c) Mean Squared Error or Mean Absolute Error Explanation: For regression tasks, decision trees use Mean Squared Error or Mean Absolute Error to evaluate splits. Q10: To improve the prediction accuracy of a decision tree for revenue prediction, which approach should you consider? a) Use a deeper tree. b) Switch to a classification algorithm. c) Use Gradient Boosting (e.g., XGBoost). d) Use a confusion matrix to optimize. Answer: c) Use Gradient Boosting (e.g., XGBoost). Explanation: Gradient Boosting combines multiple weak learners (decision trees) to improve the overall prediction accuracy, especially for regression tasks. Q11: If you suspect that irrelevant features are reducing model performance, what should you do? a) Prune the decision tree. b) Apply feature selection methods. c) Normalize all features. d) Increase the maximum depth of the tree. Answer: b) Apply feature selection methods. Explanation: Feature selection reduces noise by removing irrelevant or less important features, improving model performance. 1. What is the main objective of a classification model in machine learning? A. Predicting a continuous value B. Classifying data into predefined categories C. Clustering similar data points D. Reducing the dimensionality of the dataset Answer: B. Classifying data into predefined categories 2. Which of the following algorithms is NOT commonly used for classification? A. Logistic Regression B. Decision Tree C. K-Nearest Neighbors (KNN) D. Linear Regression Answer: D. Linear Regression 3. What does the confusion matrix in a classification problem represent? A. The relationship between input features B. A summary of the performance of a classification model C. The optimal hyperparameters for a model D. The correlation between predicted and actual continuous values Answer: B. A summary of the performance of a classification model 4. In classification problems, which metric is used to evaluate a model's ability to correctly identify positive instances? A. Precision B. Recall C. Mean Squared Error D. Root Mean Squared Error Answer: B. Recall 5. Which of the following is an example of a binary classification problem? A. Predicting the price of a house B. Classifying email as spam or not spam C. Grouping customers based on purchasing behavior D. Clustering images based on similarity Answer: B. Classifying email as spam or not spam 6. What is the role of a decision boundary in classification? A. Reducing the complexity of the model B. Separating data points into different classes C. Optimizing the loss function D. Determining the number of classes in the dataset Answer: B. Separating data points into different classes 7. Which of the following methods can be used to handle class imbalance in classification problems? A. Cross-validation B. Feature scaling C. Synthetic Minority Oversampling Technique (SMOTE) D. Gradient Descent Answer: C. Synthetic Minority Oversampling Technique (SMOTE) 8. Which algorithm uses hyperplanes to classify data points into different classes? A. Random Forest B. Naive Bayes C. Support Vector Machines (SVM) D. K-Nearest Neighbors (KNN) Answer: C. Support Vector Machines (SVM) 9. What type of activation function is commonly used in the output layer of a binary classification neural network? A. ReLU B. Softmax C. Sigmoid D. Tanh Answer: C. Sigmoid 10. In a classification task, what does an AUC-ROC score of 1 indicate? A. The model has a 50% chance of making the correct prediction B. The model performs perfectly in distinguishing between classes C. The model makes no distinction between classes D. The model requires retraining Answer: B. The model performs perfectly in distinguishing between classes 11. What is overfitting in the context of classification? A. When the model performs better on training data than unseen data B. When the model underestimates the complexity of the data C. When the model performs poorly on both training and test data D. When the model selects irrelevant features Answer: A. When the model performs better on training data than unseen data 12. Which type of classification problem involves more than two classes without a specific order? A. Binary Classification B. Multi-Class Classification C. Ordinal Classification D. Regression Answer: B. Multi-Class Classification 13. Which machine learning algorithm is best suited for high-dimensional datasets in classification tasks? A. Logistic Regression B. Random Forest C. Support Vector Machines (SVM) D. Naive Bayes Answer: C. Support Vector Machines (SVM) 14. Which evaluation metric is particularly useful when dealing with imbalanced datasets? A. Accuracy B. Precision-Recall Curve C. Mean Absolute Error D. R-squared Answer: B. Precision-Recall Curve 15. Which technique can improve classification accuracy by combining multiple models? A. Feature Scaling B. Principal Component Analysis (PCA) C. Ensemble Learning D. Stochastic Gradient Descent Answer: C. Ensemble Learning 1. Scenario You are analyzing GPS data from delivery vehicles to identify clusters of delivery locations. You notice that some areas have very dense delivery points, while others are sparse. Which of the following best explains why DBSCAN is a suitable algorithm for this task? A. It assumes clusters are spherical and of similar sizes. B. It can handle clusters of varying shapes and densities. C. It uses centroid-based clustering like K-Means. D. It requires the exact number of clusters as input. Answer: B. It can handle clusters of varying shapes and densities. 2. Scenario You are using DBSCAN to cluster customer purchase data based on transaction frequency and total spending. After running the algorithm, some data points are labeled as noise. What does this mean? A. These points belong to a smaller cluster. B. These points do not meet the density requirements for any cluster. C. These points were ignored due to missing values. D. These points were removed during preprocessing. Answer: B. These points do not meet the density requirements for any cluster. 3. Scenario A researcher is clustering data using DBSCAN. The dataset has two dense regions far apart, and there are some scattered points between them. What will DBSCAN likely do with the scattered points? A. Assign them to the nearest dense region. B. Treat them as noise if they don't satisfy density conditions. C. Merge the two dense regions into a single cluster. D. Create a new cluster for the scattered points. Answer: B. Treat them as noise if they don't satisfy density conditions. 4. Scenario You apply DBSCAN to classify satellite images based on vegetation density. The algorithm uses two parameters: epsilon (ε) and minimum points (MinPts). What will happen if you set ε too small? A. The algorithm will find fewer, larger clusters. B. Most points will be classified as noise. C. The algorithm will fail to converge. D. It will combine clusters that should be separate. Answer: B. Most points will be classified as noise. 5. Scenario In a retail store, you analyze customer footfall using DBSCAN. Two clusters represent high- density customer zones, but there are overlapping areas between the clusters. What is DBSCAN’s approach to handling overlapping clusters? A. It assigns overlapping points to both clusters. B. It randomly assigns overlapping points to one cluster. C. It assigns overlapping points to the first cluster found. D. It doesn't assign overlapping points, treating them as part of the denser cluster. Answer: D. It doesn't assign overlapping points, treating them as part of the denser cluster. 6. Scenario An environmental scientist is clustering air pollution levels in a city using DBSCAN. She notices that clusters vary in shape and density. Why is DBSCAN more effective than K- Means in this case? A. It assumes all clusters have the same density. B. It can detect clusters of arbitrary shapes. C. It requires fewer iterations to converge. D. It normalizes the data automatically. Answer: B. It can detect clusters of arbitrary shapes. 7. Scenario You are using DBSCAN to group website visitors based on browsing behavior. After running the algorithm, you observe a large number of clusters and noise points. What could be a potential solution to reduce noise? A. Increase the number of clusters. B. Increase the ε value or decrease MinPts. C. Use hierarchical clustering instead. D. Reduce the dataset size. Answer: B. Increase the ε value or decrease MinPts. 8. Scenario A geologist is clustering earthquake epicenters. The dataset contains several dense regions and outliers far from these regions. Which of the following correctly describes how DBSCAN handles outliers? A. DBSCAN ignores outliers during clustering. B. Outliers are assigned to the nearest cluster. C. Outliers are flagged as noise and not included in any cluster. D. DBSCAN merges outliers into a separate cluster. Answer: C. Outliers are flagged as noise and not included in any cluster. 9. Scenario In a dataset of customer locations, DBSCAN produces one large cluster spanning multiple regions when you increase ε. What does this indicate? A. The dataset contains too many features. B. Clusters are too dense and indistinguishable. C. The value of ε is too high, causing unrelated points to merge. D. MinPts is too low, leading to overfitting. Answer: C. The value of ε is too high, causing unrelated points to merge. 10. Scenario You are applying DBSCAN to a multidimensional dataset with unknown density distribution. Why might it be challenging to select the optimal ε? A. DBSCAN is not suitable for multidimensional data. B. Different dimensions may have varying density scales. C. The algorithm requires scaling only categorical data. D. The algorithm doesn't use ε for clustering. Answer: B. Different dimensions may have varying density scales. 1. Question What is the True Positive (TP) value in this confusion matrix? A. 50 B. 10 C. 5 D. 100 Answer: D. 100 2. Question What is the False Positive (FP) value in this confusion matrix? A. 5 B. 10 C. 100 D. 50 Answer: B. 10 3. Question What is the True Negative (TN) value in this confusion matrix? A. 50 B. 10 C. 100 D. 5 Answer: A. 50 4. Question What is the False Negative (FN) value in this confusion matrix? A. 50 B. 10 C. 5 D. 100 Answer: C. 5 5. Question What is the accuracy of the model based on the confusion matrix? Accuracy=TP + TNTotal\text{Accuracy} = \frac{\text{TP + TN}}{\text{Total}}Accuracy=TotalTP + TN A. 0.91 (91%) B. 0.88 (88%) C. 0.85 (85%) D. 0.94 (94%) Answer: A. 0.91 (91%) 6. Question What is the precision for the "YES" class? Precision=TPTP + FP\text{Precision} = \frac{\text{TP}}{\text{TP + FP}}Precision=TP + FPTP A. 0.80 (80%) B. 0.90 (90%) C. 0.91 (91%) D. 0.95 (95%) Answer: D. 0.95 (95%) 7. Question What is the recall for the "YES" class? Recall=TPTP + FN\text{Recall} = \frac{\text{TP}}{\text{TP + FN}}Recall=TP + FNTP A. 0.91 (91%) B. 0.95 (95%) C. 0.94 (94%) D. 0.96 (96%) Answer: C. 0.94 (94%) 8. Question What is the F1-score for the "YES" class? F1-score=2×Precision×RecallPrecision+Recall\text{F1-score} = 2 \times \frac{\text{Precision} \times \text{Recall}}{\text{Precision} + \text{Recall}}F1- score=2×Precision+RecallPrecision×Recall A. 0.92 (92%) B. 0.94 (94%) C. 0.95 (95%) D. 0.96 (96%) Answer: B. 0.94 (94%) 9. Question If the model predicts "NO," how often is it correct? This refers to the Negative Predictive Value (NPV). NPV=TNTN + FN\text{NPV} = \frac{\text{TN}}{\text{TN + FN}}NPV=TN + FNTN A. 0.90 (90%) B. 0.91 (91%) C. 0.93 (93%) D. 0.95 (95%) Answer: C. 0.91 (91%) 10. Question What type of error does a False Positive (FP) represent in this context? A. The model predicts "YES" when the actual is "NO." B. The model predicts "NO" when the actual is "YES." C. The model correctly predicts "NO." D. The model correctly predicts "YES." Answer: A. The model predicts "YES" when the actual is "NO."