CAIP DOMAIN 3 EXAM QUESTIONS
WITH 100% CORRECT ANSWERS
What is accuracy, and when can it be misleading in classification tasks? - Answer-
Accuracy measures correct predictions but can be misleading in imbalanced datasets.
How does precision differ from recall in classification metrics? - Answer-Precision
measures correct positive predictions, recall measures correct identification of actual
positives.
What is the F1 score, and when is it useful? - Answer-The F1 score balances precision
and recall, useful in imbalanced datasets.
When should you use the precision-recall curve over the ROC curve? - Answer-Use the
precision-recall curve for imbalanced datasets where the minority class is critical.
How is mean squared error (MSE) different from mean absolute error (MAE) in
regression tasks? - Answer-MSE emphasizes larger errors by squaring them, while
MAE treats all errors equally.
What is the purpose of cost-sensitive evaluation in machine learning? - Answer-Cost-
sensitive evaluation assigns different costs to errors based on business goals.
What is a cost matrix, and how is it used in model evaluation? - Answer-A cost matrix
assigns costs to false positives, false negatives, true positives, and true negatives.
What is threshold tuning, and why is it important in fraud detection? - Answer-Threshold
tuning adjusts decision thresholds to balance false positives and false negatives.
How does the area under the ROC curve (AUC) help evaluate model performance? -
Answer-AUC measures the model's ability to distinguish between classes.
What is the difference between regression and classification performance metrics? -
Answer-Regression metrics evaluate numerical predictions, classification metrics
evaluate discrete predictions.
What does a large gap between training and validation error indicate? - Answer-It
indicates overfitting.
,How does high bias affect model performance? - Answer-High bias leads to underfitting,
where the model fails to capture complex patterns.
What is the effect of high variance on a model's performance? - Answer-High variance
leads to overfitting, where the model memorizes noise in the data.
How can you reduce bias in a model? - Answer-Increase model complexity or add more
features.
What regularization techniques help reduce model variance? - Answer-L1 and L2
regularization.
What is overfitting, and how can it be prevented? - Answer-Overfitting is when a model
learns noise in the training data; prevent it with regularization, dropout, or cross-
validation.
What are the three main types of machine learning algorithms? - Answer-Supervised
learning, unsupervised learning, reinforcement learning.
What is the difference between supervised and unsupervised learning? - Answer-
Supervised learning uses labeled data, while unsupervised learning works with
unlabeled data.
Which machine learning algorithm is best suited for classification tasks? - Answer-
Support Vector Machines (SVM) and Logistic Regression.
What is a key use case for reinforcement learning? - Answer-Autonomous driving or
robotics.
How does supervised learning work? - Answer-It uses labeled data to map inputs to
known outputs.
What type of data does unsupervised learning work with? - Answer-Unlabeled data to
uncover patterns and relationships.
What is Q-learning used for in reinforcement learning? - Answer-It helps an agent learn
an optimal policy to maximize cumulative rewards.
Which algorithm is used for clustering tasks? - Answer-K-Means Clustering.
What are some key challenges in reinforcement learning? - Answer-Balancing
exploration vs. exploitation and data efficiency.
Why is data availability important in supervised learning? - Answer-Supervised learning
requires large amounts of labeled data for training.
, What is the key feature of Convolutional Neural Networks (CNNs)? - Answer-CNNs are
designed to process image and visual data by learning hierarchical features.
What are Recurrent Neural Networks (RNNs) designed for? - Answer-RNNs are
designed for sequential data, such as time-series and language processing.
What problem do LSTMs solve in sequential data processing? - Answer-LSTMs solve
the vanishing gradient problem, allowing networks to capture long-term dependencies.
What is the primary function of Transformers in NLP? - Answer-Transformers process
sequential data using attention mechanisms without step-by-step recurrence.
Why are CNNs effective in image classification tasks? - Answer-CNNs capture spatial
relationships and patterns in pixel data, making them ideal for image recognition.
How do RNNs handle sequential data? - Answer-RNNs retain memory of previous
inputs, making them useful for tasks like text and speech analysis.
What applications are LSTMs commonly used for? - Answer-LSTMs are commonly
used for machine translation, text generation, and speech recognition.
How do Transformers improve upon traditional RNNs? - Answer-Transformers use
parallel processing, which is more efficient than the sequential nature of RNNs.
What is the purpose of the self-attention mechanism in Transformers? - Answer-The
self-attention mechanism allows the model to weigh the importance of different parts of
the sequence.
What makes CNNs suitable for visual data analysis? - Answer-They automatically learn
relevant features from visual data such as images.
What is the purpose of Principal Component Analysis (PCA)? - Answer-PCA reduces
the dimensionality of data while preserving the most important variance.
How is feature importance used in predictive modeling? - Answer-Feature importance
helps identify which variables contribute most to the model's predictions.
What is the role of autoencoders in feature extraction? - Answer-Autoencoders
compress input data into a lower-dimensional space, capturing key features.
How are linear models used in pattern recognition? - Answer-Linear models are used
for tasks where relationships between inputs and outputs are linear.
What are non-linear models better suited for? - Answer-Non-linear models, like Decision
Trees, are suited for capturing complex relationships.
WITH 100% CORRECT ANSWERS
What is accuracy, and when can it be misleading in classification tasks? - Answer-
Accuracy measures correct predictions but can be misleading in imbalanced datasets.
How does precision differ from recall in classification metrics? - Answer-Precision
measures correct positive predictions, recall measures correct identification of actual
positives.
What is the F1 score, and when is it useful? - Answer-The F1 score balances precision
and recall, useful in imbalanced datasets.
When should you use the precision-recall curve over the ROC curve? - Answer-Use the
precision-recall curve for imbalanced datasets where the minority class is critical.
How is mean squared error (MSE) different from mean absolute error (MAE) in
regression tasks? - Answer-MSE emphasizes larger errors by squaring them, while
MAE treats all errors equally.
What is the purpose of cost-sensitive evaluation in machine learning? - Answer-Cost-
sensitive evaluation assigns different costs to errors based on business goals.
What is a cost matrix, and how is it used in model evaluation? - Answer-A cost matrix
assigns costs to false positives, false negatives, true positives, and true negatives.
What is threshold tuning, and why is it important in fraud detection? - Answer-Threshold
tuning adjusts decision thresholds to balance false positives and false negatives.
How does the area under the ROC curve (AUC) help evaluate model performance? -
Answer-AUC measures the model's ability to distinguish between classes.
What is the difference between regression and classification performance metrics? -
Answer-Regression metrics evaluate numerical predictions, classification metrics
evaluate discrete predictions.
What does a large gap between training and validation error indicate? - Answer-It
indicates overfitting.
,How does high bias affect model performance? - Answer-High bias leads to underfitting,
where the model fails to capture complex patterns.
What is the effect of high variance on a model's performance? - Answer-High variance
leads to overfitting, where the model memorizes noise in the data.
How can you reduce bias in a model? - Answer-Increase model complexity or add more
features.
What regularization techniques help reduce model variance? - Answer-L1 and L2
regularization.
What is overfitting, and how can it be prevented? - Answer-Overfitting is when a model
learns noise in the training data; prevent it with regularization, dropout, or cross-
validation.
What are the three main types of machine learning algorithms? - Answer-Supervised
learning, unsupervised learning, reinforcement learning.
What is the difference between supervised and unsupervised learning? - Answer-
Supervised learning uses labeled data, while unsupervised learning works with
unlabeled data.
Which machine learning algorithm is best suited for classification tasks? - Answer-
Support Vector Machines (SVM) and Logistic Regression.
What is a key use case for reinforcement learning? - Answer-Autonomous driving or
robotics.
How does supervised learning work? - Answer-It uses labeled data to map inputs to
known outputs.
What type of data does unsupervised learning work with? - Answer-Unlabeled data to
uncover patterns and relationships.
What is Q-learning used for in reinforcement learning? - Answer-It helps an agent learn
an optimal policy to maximize cumulative rewards.
Which algorithm is used for clustering tasks? - Answer-K-Means Clustering.
What are some key challenges in reinforcement learning? - Answer-Balancing
exploration vs. exploitation and data efficiency.
Why is data availability important in supervised learning? - Answer-Supervised learning
requires large amounts of labeled data for training.
, What is the key feature of Convolutional Neural Networks (CNNs)? - Answer-CNNs are
designed to process image and visual data by learning hierarchical features.
What are Recurrent Neural Networks (RNNs) designed for? - Answer-RNNs are
designed for sequential data, such as time-series and language processing.
What problem do LSTMs solve in sequential data processing? - Answer-LSTMs solve
the vanishing gradient problem, allowing networks to capture long-term dependencies.
What is the primary function of Transformers in NLP? - Answer-Transformers process
sequential data using attention mechanisms without step-by-step recurrence.
Why are CNNs effective in image classification tasks? - Answer-CNNs capture spatial
relationships and patterns in pixel data, making them ideal for image recognition.
How do RNNs handle sequential data? - Answer-RNNs retain memory of previous
inputs, making them useful for tasks like text and speech analysis.
What applications are LSTMs commonly used for? - Answer-LSTMs are commonly
used for machine translation, text generation, and speech recognition.
How do Transformers improve upon traditional RNNs? - Answer-Transformers use
parallel processing, which is more efficient than the sequential nature of RNNs.
What is the purpose of the self-attention mechanism in Transformers? - Answer-The
self-attention mechanism allows the model to weigh the importance of different parts of
the sequence.
What makes CNNs suitable for visual data analysis? - Answer-They automatically learn
relevant features from visual data such as images.
What is the purpose of Principal Component Analysis (PCA)? - Answer-PCA reduces
the dimensionality of data while preserving the most important variance.
How is feature importance used in predictive modeling? - Answer-Feature importance
helps identify which variables contribute most to the model's predictions.
What is the role of autoencoders in feature extraction? - Answer-Autoencoders
compress input data into a lower-dimensional space, capturing key features.
How are linear models used in pattern recognition? - Answer-Linear models are used
for tasks where relationships between inputs and outputs are linear.
What are non-linear models better suited for? - Answer-Non-linear models, like Decision
Trees, are suited for capturing complex relationships.
