Font size
WorksheetsBUSA3020 Quiz W7-W12
Total questions: 93
Worksheet time: 47mins
What are pipelines in scikit-learn used for?
None of the listed options
Combining multiple classifiers models into one
Combining multiple processing steps into one
Visualising data transformation steps
Generating synthetic data samples
Which of the following is a primary advantage of using pipelines in scikit-learn?
Automatic data collection
Eliminating the need for data validation and testing
None of the listed options
Simplicity and Convenience
Reducing the need for data cleaning
How do pipelines improve the reproducibility of machine learning workflows?
By increasing the accuracy of predictions
By reducing the complexity of algorithms
None of the listed options
By automatically updating data sources
By making preprocessing steps explicit
What is the role of the fit method in a scikit-learn pipeline?
To predict new data points
To visualise the data transformation process
To split the data into training and testing sets
To pass data through transformers and train the estimator
None of the listed options
What does the predict method in a scikit-learn pipeline do?
Determines the best hyperparameters for the model
None of the listed options
Calculates the accuracy of the model
Do intermediate steps and predict using the final estimator
Fits the data to the pipeline
What does the holdout method in machine learning involve?
Only using a training set to evaluate the model
Using the entire dataset for both training and testing
Evaluating the model using cross-validation only
Split of the dataset into training, validation, and test sets
None of the listed options
K-fold cross-validation improves upon the holdout method by
Using each fold as a validation set exactly once
Only using one fold for both training and testing
None of the listed options
Eliminating the need for a validation set
Reducing the computational complexity of model evaluation
The need for cross-validation arises due to
The high cost of data collection
The simplicity of machine learning models
None of the listed options
The availability of large datasets
Limitations of the holdout method
What is the purpose of hyperparameter tuning via grid search?
To reduce the size of the dataset
To increase the speed of the training process
To select the best machine learning algorithm
None of the listed options
To find the best combination of hyperparameters values
What do learning curves plot?
Model accuracy versus the sample size
Model accuracy versus the number of features
The number of iterations versus the error rate
The learning rate versus the validation accuracy
None of the listed options
What can validation curves help identify in a model?
None of the listed options
The best machine learning algorithm
The most important features
The optimal size of the dataset
Overfitting and underfitting
Which of the following is a component of a confusion matrix?
True Positive Rate
None of the listed options
False Negative Rate
Unlabeled samples
False Discoveries
Precision in the context of classification models is:
The ratio of true positives observations to the total predicted positives
The ratio of correctly predicted positive observations to the total observations
The ratio of correctly predicted negatives to the total predicted negatives
None of the listed options
The accuracy of the model
The F1 score is used to
Determine the learning rate
Calculate the model's training speed
Balance precision and recall
None of the listed options
Measure the model's accuracy only
Recall is calculated as
TP/(TP+FP)
TN/(TN+FP)
None of the listed options
TN/(TN+FN)
TP/(FN+TP)
Which of the following best defines a centroid in clustering?
The average of all points in a cluster
The most distant point in a cluster
The point with maximum variance
None of the listed options
The least central point in a cluster
A medoid is best defined as
Always positioned at the edge of a cluster
A randomly chosen point in a cluster
None of the listed options
An average point in multidimensional space
The most centrally located point within a cluster
What is the main advantage of K-Means++ over K-Means?
It spreads out the initial centroids
It selects initial centroids randomly
It produces more clusters
It requires fewer clusters
None of the listed options
The K-Means algorithm is not suited for
Assigning data points to the nearest cluster
Finding the longest distance between clusters
None of the listed options
Partitioning n items into k clusters
Minimising within-cluster variances
Euclidean distance is used to
Measure the total variance within a cluster
Calculate the density of clusters
Determine the highest point in a dataset
None of the listed options
Quantify the dissimilarity between points in space
The Elbow method is used to determine
The best distance metric for clustering
The optimal number of clusters in a dataset
The maximum number of iterations for K-Means
The number of outliers in a dataset
None of the listed options
Silhouette plots examine
None of the listed options
The number of iterations needed for convergence
The similarity of features within a cluster
The quality of clustering
The maximum distance between clusters
Hierarchical clustering does not include
Organising clusters as trees
Random assignment of clusters
None of the listed options
Agglomerative (Bottom-Up) Method
Divisive (Top-Down) Method
The complete linkage method
Defines the distance between two clusters as the maximum distance between any single data point in the first cluster and any single data point in the second cluster
Calculates the distance between clusters based on the shortest distance from any member of one cluster to any member of the other cluster
Merges clusters based on the average distance between all pairs of points in the two clusters
Is a type of agglomerative clustering
Defines the distance between two clusters as the distance between their centroids
A dendrogram is used in
None of the listed options
Only divisive clustering
Only K-Means clustering
Illustrating the arrangement of hierarchical clustering
Displaying the exact coordinates of data points
DBSCAN algorithm classifies clusters based on
The number of centroids
None of the listed options
The density of data points
The similarity of feature-target variable pairs
The maximum distance between any two points
Which of the following is not a key concept of DBSCAN?
None of the listed options
Noise Point
Core Point
Border Point
Centroid Point
What does the initial step in K-Means clustering involve?
Evaluating the model's accuracy
Selecting k centroids
Calculating the cluster count
Assigning data points to existing clusters
None of the listed options
Which of the following is not a goal of unsupervised learning?
To predict the outcome for new, unseen data
To find clusters or associations without knowing the true labels
To explore the structure of the data to find patterns
None of the listed options
To model the underlying structure or distribution in the data
In the context of Euclidean distance, what is d(y,x) if y and x are vectors?
The maximum distance between two points in a dataset
The product of two vectors y and x
The average position of all points in a cluster
None of the listed options
The length of the line segment connecting x and y
What distinguishes the Elbow method in cluster analysis?
None of the listed options
It identifies the point where increasing the number of clusters significantly improves the fit of the data
It calculates the number of points in each cluster
It determines the minimum Euclidean distance for optimal clustering
It identifies the point where increasing the number of clusters does not significantly improve the fitting of the data
Which statement about DBSCAN is false?
It requires pre-specifying the number of clusters
None of the listed options
It classifies points as core points, border points, or noise points
Core points have a minimum number of points within a given radius
It is based on the density of data points
What does an ensemble method involve in machine learning?
Focusing solely on reducing variance in predictions
Ensemble methods are designed to
Using only different types of models within model sets
Generating multiple models and combining their predictions
None of the listed options
Ensemble methods are designed to
Reduce only bias
None of the listed options
Reduce only variance
Increase both bias and variance
Focus solely on model accuracy
What is majority voting in the context of ensemble methods?
When more than half of the classifiers agree on the same class
When each classifier has one vote, regardless of confidence
None of the listed options
When only one classifier decides the final prediction
When a class needs 100% of the votes to be chosen
Plurality voting is used when
There's only one possible outcome
The class with more than 50% of the wins
None of the listed options
The class with the most votes wins
A class receives exactly half of the votes
Hard voting in predictive ensembles is characterised by
Ignoring the votes from less accurate classifiers
Considering only the classifiers' confidence levels
Summing the weighted votes for each class label
None of the listed options
Averaging probability estimates from classifiers
Soft voting differs from hard voting by
Requiring unanimous agreement among classifiers
Using only the most confident classifier's prediction
None of the listed options
Averaging the probability estimates for each class label
Not using votes from classifiers
Bagging is an ensemble method that
Can reduce bias without significantly increasing variance
None of the listed options
Increases variance to avoid overfitting
Can reduce variance without significantly increasing bias
Trains each model on the entire dataset without replacement
A weak learner is a model that
Performs worse than random guessing
None of the listed options
Is always a model with high bias
Is always a model with high variance
Cannot be combined to improve performance
Adaptive Boosting (AdaBoost) works by
Reducing variance while increasing bias
Combining strong learners into a single model
Focusing solely on correctly classified instance
None of the listed options
Combining multiple weak learners to form a strong learner
In AdaBoost, the weights of instances are updated to
None of the listed options
Decrease emphasis on misclassified instances
Give more emphasis to misclassified instances
Ensure all instances have equal weight
Give more emphasis to correctly classified instances
AdaBoost is sensitive to noisy data and outliers because
It focuses on instances that are hard to classify
It assigns lower weights to them automatically
It ignores them during the training process
It uses them to reduce the model's bias
None of the listed options
The main difference between Bagging and AdaBoost is that Bagging
None of the listed options
Is sensitive to noisy data and outliers
Reduces bias only
Creates a set of dependent classifiers
Reduces variance without significantly affecting bias
AdaBoost updates the instance weights by
Randomly assigning weights in each iteration
Keeping the weights constant throughout the learning process
None of the listed options
Increasing weights for misclassified instances
Decreasing weights for instances hard to classify
What is the role of weights in soft voting?
They have no role in soft voting
They are used to average the probability estimates for class labels
They decide which classifiers to exclude from the ensemble
None of the listed options
They determine the final prediction through majority voting
Which statement is not true about majority voting in ensemble methods?
It applies to binary classification problems
It always leads to correct classification
It is a method to aggregate predictions from multiple models
None of the listed options
It requires more than 50% of the votes for a decision in binary classification
Which of the following is not true about soft voting?
It averages probability estimates for each class label
It relies on the confidence levels of classifiers' predictions
It considers only the class labels predicted by each classifier
None of the listed options
It can involve weighted averaging of probabilities
Which statement is NOT true regarding AdaBoost?
None of the listed options
It can be sensitive to noise and outliers
It involves an iterative reweighting of instances
It focuses more on instances that are harder to classify
It combines multiple strong learners to reduce bias
Which of the following is not true?
Simple regression requires more data
Multiple regression is used when there are no confounding variables
All of the listed options
Simple regression is used for predicting multiple outputs
Multiple regression is computationally less intensive
What of the following techniques is used in Exploratory Data Analysis (EDA)?
Inferential statistics
Descriptive statistics
Predictive modelling
None of the listed options
Validation analysis
What do scatter plots primarily show?
The relationship between two variables
The spread/variance of a univariate variable
None of the listed options
The central tendency of a univariate variable
The correlation coefficient between two variables
What does a correlation coefficient of zero indicate?
Perfect negative linear relationship
No relationship
Perfect positive linear relationship
No linear relationship
High degree of correlation
How is the correlation matrix useful in multiple regression analysis?
Can aid in building a regression model
Helps detect multicollinearity
Visualising data connections
All of the listed options
Feature selection
Why is the RANSAC algorithm preferred in the presence of data anomalies?
None of the listed options
It uses less data
It reduces computational time
It is robust to outliers
It handles multiple outputs
Which method is not used to evaluate the performance of a linear regression model?
Residual Plots
None of the listed options
Correlations
R-squared (R²)
Mean Squared Error (MSE)
What does Lasso Regression do differently from Ridge Regression?
Uses squared coefficients in the penalty
None of the listed options
Shrinks more coefficients to zero
Uses a combination of L1 and L2 penalties
Is computationally efficient
Which is a potential disadvantage of polynomial regression models?
None of the listed options
Risk of overfitting
Cannot handle interactive terms
Cannot model linear relationships
Too simple for complex datasets
What is an advantage of Random Forest Regression over single Decision Tree Regression?
Random Forests are not linear
Easier to compute
Requires less data
Less likely to overfit
None of the listed options
Which regression method includes both L1 and L2 regularisation techniques?
Quadratic Polynomial Regression
None of the listed options
Elastic Net Regression
Lasso Regression
Ridge Regression
In the context of regression models, what does RANSAC stand for?
Randomised Statistical Consensus
Random Sample Consensus
Regression Analysis Sample Consensus
None of the listed options
Regression Sample Consensus
Which metric is specifically used to assess the goodness of fit in regression models?
Mean Squared Error (MSE)
Mean Absolute Percentage Error (MAPE)
None of the listed options
Root Mean Squared Error (RMSE)
Median Absolute Absolute (MAE)
Which of the following is true about Random Forest and Gradient Boosting ensemble methods?
Both methods can be only used for classification problems
None of the listed options
Random Forest is used for classification whereas Gradient Boosting is used for regression problems
Both methods can be only used for regression problems
Random Forest is used for regression problems whereas Gradient Boosting is used for Classification problems
The results below are obtained for three different models based on the same dataset. Which model exhibits the overfitting problem?
Model-1
MSE based on training data = 17.19
MSE based on testing data = 20.20
Model-2
MSE based on training data = 39.20
MSE based on testing data = 63.50
Model-3
MSE based on training data = 68.3
MSE based on testing data = 69.4
None of the listed models
Model 2
Model 3
Model 1
All of the listed models
Which of the following best describes time series data?
Data collected over time at regular intervals
Data where observations are always independent of each other
Data collected at a single point in time
Data with no temporal component
None of the listed options
Which type of model is appropriate when the current value of a time series depends only on previous values of the series?
Last historical values of the series
None of the listed options
Autoregressive
White Noise
Moving Average
Which of the following best describes white noise?
No temporal dependence
None of the listed options
High autocorrelation at all lags
Non-zero mean
Positive trend
The ACF (Autocorrelation Function) of a white noise process is:
Zero for all non-zero lags
Gradually decaying towards zero
None of the listed options
Zero for all lags
Non-zero for all lags
In a Moving Average (MA) model, the current value of the time series depends on:
Past error terms
Future values of the series
Past values of the series
None of the listed options
Only the most recent observation
Which of the following models combines both AR and MA components?
Autoregressive Moving Average Model
White Noise
Vector Autoregressive Model
None of the listed options
Generalised Autoregressive Model
If the PACF of a time series cuts off after lag 2 while ACF decays gradually, what model is most likely appropriate?
None of the listed options
AR(2)
AR(1)
ARMA(2,1)
MA(2)
Which of the following statements is true for an AR(1) process?
It depends on the previous value and error term
It depends only on the current error term
None of the listed options
It is a white noise process by definition
It depends on the previous and current error terms
A time series model where the current value depends only on current and past error terms is known as:
White Noise
Moving Average (MA) model
Cross-sectional model
Autoregressive (AR) model
None of the listed options
What does the autocorrelation function (ACF) measure in a time series?
The frequency of seasonal patterns
The correlation between lagged values and current values
The relationship between current and future values
The standard deviation of the time series
None of the listed options
In a time series, seasonality refers to:
None of the listed options
A long-term upward or downward movement in the data
A temporary shock to the series
Regular patterns that repeat at fixed intervals
Random fluctuations with no pattern
Which of the following is true about the Moving Average (MA) model?
The ACF cuts off after lag q
Both ACF and PACF cut off after q lags
None of the listed options
The PACF cuts off after lag q
The ACF gradually decays
Which of the following is true about the residuals of a well-fitted ARMA model?
The residuals should behave like white noise
The residuals should have increasing variance
The residuals should exhibit high autocorrelation
None of the listed options
The residuals should show a trend over time
In a well-fitted ARMA model, the ACF and PACF of the residuals should:
None of the listed options
None of the listed options
Show a gradual decay
Show significant autocorrelation
Should cut off to zero after lag 3
Be zero for all non-zero lags
What does the Partial Autocorrelation Function (PACF) measure?
None of the listed options
The correlation between the current value and future values of the time series
The correlation between the current value and all previous values of the time series
The overall trend in the time series data
The correlation between the current value and lagged values, after removing the effects of intermediate lags
What is Sentiment Analysis primarily used for in the context of machine learning?
Evaluating the grammatical accuracy of text
None of the listed options
Translating text from one language to another
Converting speech into text
Predicting the next word from textual data
Which application of Sentiment Analysis was mentioned in the lecture notes?
Analysing email sentiments for marketing purposes
Sorting product reviews into predefined categories
None of the listed options
Distinguishing between positive and negative movie reviews on IMDb
Building a predictor for financial market trends based on news articles
What does the Bag-of-Words model ignore when processing text?
The frequency of words
The uniqueness of words
The grammar and order of words
None of the listed options
All of the listed options
Which is not a core component of the Bag-of-Words model?
Vocabulary Creation
Tokenization
Syntax Parsing
None of the listed options
Vectorization
What does TF-IDF stand for?
Term Formulation-Inverse Document Frequency
Total Frequency-Integrated Document Frequency
None of the listed options
Term Frequency-Inverse Document Frequency
Text Frequency-Inverse Document Frequency
What is an N-gram model used to represent?
A predictive model for text classification on n words
The frequency of n terms across documents
None of the listed options
The importance of n documents
A sequence of n items from text
What are regular expressions used for in Python?
String search and manipulation
Creating secure passwords
Developing web applications
Calculating numerical statistics
None of the listed options
What is a token in NLP?
None of the listed options
A regex expression in Python
A type of punctuation mark
A semantic unit of analysis
A method for text summarization
What does word stemming achieve?
None of the listed options
Reducing words to their root form
Correcting grammatical errors
Encrypting text data
Expanding the vocabulary of a text
What is the main difference between CountVectorizer and HashingVectorizer?
CountVectorizer uses Python regex; HashingVectorizer does not
HashingVectorizer is used only for large text bodies
CountVectorizer maintains a full dictionary; HashingVectorizer does not
CountVectorizer supports n-grams; HashingVectorizer does not
None of the listed options
What is Topic Modeling used for?
Encoding text data
Discovering sentiment in a collection of texts
Discovering abstract themes in texts
Correcting syntactical errors
None of the listed options
Which technique is known as out-of-core learning?
A method that involves fitting the model to all data
A technique used only in neural network training
A method that involves continuous model training
Incrementally fitting a classifier using batches of data
What is primarily reduced by the TF-IDF calculation?
The influence of common words
The need for manual review of documents
None of the listed options
The accuracy of document classification
The size of the text corpus
Which of the following is NOT true about the Bag-of-Words model?
It considers the semantic meaning of words
It maintains word frequency
It breaks text down into a collection of words
It involves vectorization of documents
None of the listed options
