Exploring Machine Learning in BI

What is a regression problem in machine learning?

• A regression problem is used for classifying discrete categories.

• A regression problem is a type of machine learning task that involves predicting continuous values.

• A regression problem predicts binary outcomes only.

• A regression problem focuses on clustering data points.

Answer explanation

A regression problem is specifically focused on predicting continuous values, making the second choice correct. The other options incorrectly describe classification, binary outcomes, or clustering tasks.

How do you evaluate the performance of a regression model?

• Evaluate based on the number of features used in the model.

• Use metrics like MAE, MSE, RMSE, and R-squared to evaluate performance.

• Check the training time of the model as the main metric.

• Use only visual inspection of the model's predictions.

Answer explanation

The correct way to evaluate a regression model's performance is by using metrics like MAE, MSE, RMSE, and R-squared, as they provide quantitative measures of prediction accuracy, unlike the other options.

What is the formula for linear regression?

• y = m + bx

• y = mx + b

• y = mx^2 + b

• y = ax^2 + c

Answer explanation

The correct formula for linear regression is y = mx + b, where m is the slope and b is the y-intercept. This equation represents a straight line, unlike the other options which involve quadratic terms.

What distinguishes classification problems from regression problems?

• Classification problems predict numerical values; regression problems predict categories.

• Classification problems involve time series; regression problems involve static data.

• Classification problems predict categories; regression problems predict continuous values.

• Classification problems require labeled data; regression problems require unlabeled data.

Answer explanation

Classification problems predict categories, such as labels or classes, while regression problems predict continuous values, like real numbers. This distinction is fundamental in machine learning tasks.

What metrics are commonly used to evaluate classification models?

• Log Loss

• R-squared

• Mean Squared Error

• Accuracy, Precision, Recall, F1 Score, ROC-AUC, Confusion Matrix

Answer explanation

The correct choice includes metrics specifically for evaluating classification models, such as Accuracy, Precision, Recall, F1 Score, ROC-AUC, and Confusion Matrix. Log Loss is for probabilistic models, while R-squared and Mean Squared Error are for regression.

Explain the concept of Bayesian methods in classification.

• Bayesian methods rely solely on prior probabilities without considering new evidence.

• Bayesian methods use Bayes' theorem to update class probabilities based on evidence.

• Bayesian methods are only applicable to linear classification problems.

• Bayesian methods do not involve any statistical calculations.

Answer explanation

Bayesian methods utilize Bayes' theorem to update the probability of a class based on new evidence, allowing for dynamic classification as more data becomes available. This distinguishes them from methods that rely solely on prior probabilities.

What is logistic regression and when is it used?

• Logistic regression is used for binary classification problems.

• Logistic regression is used for multi-class classification problems.

• Logistic regression is primarily used for clustering data points.

• Logistic regression is a method for regression analysis of continuous outcomes.

Answer explanation

Logistic regression is a statistical method used for binary classification problems, where the outcome is a binary variable. It estimates the probability that a given input belongs to a particular category.

What are the main types of clustering methods?

• Partitioning, Hierarchical, Density-based, Model-based

• Supervised, Unsupervised, Reinforcement

• Regression, Classification, Association

• Linear, Non-linear, Time-series

Answer explanation

The main types of clustering methods include Partitioning, Hierarchical, Density-based, and Model-based. These methods categorize data into groups based on similarity, making them essential for unsupervised learning.

Describe partition methods in clustering.

• Partition methods in clustering include k-means, k-medoids, and k-modes.

• Spectral clustering methods

• Hierarchical clustering methods

• Density-based clustering methods

Answer explanation

Partition methods in clustering, such as k-means, k-medoids, and k-modes, focus on dividing data into distinct groups based on similarity. These methods differ from spectral, hierarchical, and density-based clustering approaches.

What are hierarchical methods in clustering?

• Methods that only use linear regression for clustering

• Techniques that require labeled data for grouping

• Algorithms that do not consider data similarity in clustering

• Hierarchical methods in clustering are techniques that create a tree-like structure to group data points based on their similarities, using either agglomerative or divisive approaches.

Answer explanation

Hierarchical methods in clustering create a tree-like structure to group data points based on their similarities. They use agglomerative (bottom-up) or divisive (top-down) approaches, making them distinct from other methods.