Machine Learning Concepts Quiz

Questions and Answers

What is the primary goal of market segmentation?

Increase the size of the consumer market

Simplify product offerings for consumers

Divide the market into groups based on similar responses to marketing (correct)

Generate a single marketing message for all consumers

In matrix factorisation, what is typically true about the sizes of factor matrices U and V?

Both must be of the same size

k must be less than both n and p (correct)

U must be larger than V

k must be equal to n and p

Which of the following applications is NOT commonly associated with anomaly detection?

Manufacturing quality control

Network intrusion detection

Market segmentation (correct)

Fraud detection

What is the main objective of reinforcement learning?

Develop a mapping from states to actions to maximize rewards Signup and view all the answers

Which method is particularly effective for structured data problems in machine learning?

Gradient boosting and tree-based methods Signup and view all the answers

What characteristic defines structured data compared to unstructured data?

Structured data has a predictable format Signup and view all the answers

In the context of matrix factorisation, what is the purpose of dimensionality reduction?

To simplify the dataset while retaining essential information Signup and view all the answers

Which of the following best describes anomaly detection?

A technique for detecting significantly different data points Signup and view all the answers

Why is it necessary to scale inputs before running the kNN algorithm?

To make the Euclidean distance calculation valid Signup and view all the answers

What is a limitation of using linear regression compared to kNN?

It can only make predictions based on a linear function Signup and view all the answers

What characterizes the kNN regression method?

It is based on the similarity of examples for predictions Signup and view all the answers

Which of the following is true regarding the curse of dimensionality in kNN?

It hampers performance with many predictors Signup and view all the answers

What is an advantage of linear regression over kNN?

It has low variance and is highly interpretable Signup and view all the answers

Which statement about kNN is false?

It is highly interpretable and user-friendly Signup and view all the answers

In what way does choosing different values of k impact the kNN model?

It influences the smoothness of the predictive function Signup and view all the answers

What is a consequence of using kNN with a high number of predictors?

It can cause breakdown due to dimensionality issues Signup and view all the answers

What is the primary range that max-abs scaling focuses on?

[-1, 1] Signup and view all the answers

Which transformation would be most effective for reducing right skewness in data?

Log transformation Signup and view all the answers

What is a primary advantage of robust scaling over other scaling methods?

It performs better with outliers. Signup and view all the answers

In which situation might you want to create a dummy variable?

When a variable has many zeros. Signup and view all the answers

What does the Box-Cox transformation require as input?

A transformation parameter λ and a shift parameter α. Signup and view all the answers

What is a key characteristic of the Yeo-Johnson transformation?

It can handle both positive and negative predictors. Signup and view all the answers

How should discrete predictors with many possible values be treated?

As continuous variables. Signup and view all the answers

Why is encoding nominal variables necessary?

Algorithms often require numerical features. Signup and view all the answers

What is a common issue to identify during univariate exploratory data analysis (EDA)?

High cardinality Signup and view all the answers

Which measure of dependence is specifically used for continuous variables?

Pearson correlation Signup and view all the answers

In bivariate exploratory data analysis (EDA), which aspect indicates a potential problem with model assumptions?

Non-constant error variance Signup and view all the answers

Which of the following terms describes the process of preparing data for machine learning algorithms?

Feature engineering Signup and view all the answers

What issue should be monitored during multivariate EDA?

Outliers Signup and view all the answers

Which correlation coefficient is appropriate for analyzing ordered categorical variables?

Kendall’s τ rank correlation Signup and view all the answers

Which of the following is an indicator of multicollinearity in multivariate data analysis?

Global correlation coefficients Signup and view all the answers

What does feature engineering NOT typically involve?

Gathering additional data from external sources Signup and view all the answers

What is the first step in the k-Nearest Neighbours (kNN) algorithm when making a prediction?

Finding the k training examples closest to the test input Signup and view all the answers

What does the notation Nk(x, D) represent in kNN?

The set of indexes for the nearest neighbors Signup and view all the answers

In kNN, what is the purpose of selecting the parameter k?

To average the response values for the nearest neighbors Signup and view all the answers

If k = 1 in kNN, what will the prediction be based on?

The response value of the closest training example only Signup and view all the answers

When k = 2 in kNN, how is the prediction calculated?

By averaging the response values of the two nearest neighbors Signup and view all the answers

What type of learning method does k-Nearest Neighbours represent?

Supervised learning Signup and view all the answers

In the provided example, what is the correct response predicted by kNN with k=1 for the test input with a salary of 59100?

1006 Signup and view all the answers

Which of the following statements about kNN is true?

kNN stores all training examples in memory for future predictions Signup and view all the answers

Study Notes

Market Segmentation

Divides a diverse consumer market into groups based on preferences, requirements, and response tendencies.
Aims to enhance marketing effectiveness by targeting similar responding groups.

Matrix Factorization

Decomposes a matrix X into two factor matrices U and V, with dimensions n × k and p × k (k < min{n, p}).
Important in dimensionality reduction, simplifying datasets while preserving key information.

Anomaly Detection

Identifies data points significantly different from the rest, also known as outlier detection.
Used in fraud detection, network intrusion detection, and ensuring manufacturing quality.

Reinforcement Learning

A machine learning method where an agent makes decisions through actions in an environment, receiving rewards as feedback.
Aims to determine the optimal policy to maximize cumulative rewards over time.

k-Nearest Neighbors (kNN)

A predictive method that uses proximity to training examples in memory to predict outcomes for test inputs.
The prediction for a test input is the average response of the k closest training examples.
Scales input data before applying the algorithm for effectiveness, using Euclidean distance for measurement.

Linear Regression vs. k-Nearest Neighbors

Linear Regression: Utilizes a linear predictive function based on optimization, interpretable, quick training, generally low variance, and scales well. However, it struggles with non-linear relationships.
k-Nearest Neighbors: Highly flexible and can model complex relationships but is sensitive to the curse of dimensionality and slow for large datasets. It does not assume a functional form.

Exploratory Data Analysis (EDA)

Univariate EDA: Focuses on data errors, missing values, outliers, skewness, kurtosis, multi-modality, and high cardinality.
Bivariate EDA: Examines relationships, identifies weak/strong correlations, non-linearity, and outliers.

Feature Engineering

The process of preparing data for learning algorithms, crucial for project success.
Includes extracting, constructing, and processing features to optimize algorithm performance.

Feature Scaling

Standardization or scaling can enhance model performance; robust scaling is useful in the presence of outliers.
Log transformations can normalize data, especially for skewed distributions.

Transformations

Box-Cox Transformation: A flexible transformation that adjusts data distribution based on a parameter.
Yeo-Johnson Transformation: An extension of Box-Cox that accommodates both positive and negative values.

Handling Zeros and Discrete Predictors

Create dummy variables or treat many zero values distinctly for more effective modeling.
Discrete predictors may be treated as continuous or categorical, based on their value range.

Categorical Predictors

Nominal variables must be encoded numerically for use in machine learning models.

Studying That Suits You

Use AI to generate personalized quizzes and flashcards to suit your learning preferences.

Description

Test your knowledge on essential machine learning concepts including market segmentation, matrix factorization, anomaly detection, reinforcement learning, and k-nearest neighbors. This quiz covers key principles and applications in a concise format to enhance your understanding of the subject.