Unsupervised Learning Techniques Flashcards
6 cards from real MS-DS Master of Data science practice questions. Tap to flip, then mark Knew It or Still Learning — missed cards come back until you master them.
Read the first 6 Unsupervised Learning Techniques flashcards as text
A marketing team wants to segment its customer base into distinct groups based on purchasing behaviors (e.g., frequency, recency, monetary value) but has no predefined group labels for the customers. Which unsupervised learning technique is most suitable for this task?
Answer: K-Means Clustering
K-Means Clustering is a partitional clustering algorithm designed to group unlabeled data into a pre-specified number of clusters (K), where each data point belongs to the cluster with the nearest mean. This is ideal for customer segmentation. PCA is for dimensionality reduction, while Linear Regression and SVMs are supervised learning algorithms.
What is the primary objective of Principal Component Analysis (PCA) when used as an unsupervised learning technique?
Answer: To reduce dimensionality by finding a new set of uncorrelated variables (principal components) that capture the maximum variance in the data.
PCA is a dimensionality reduction technique. Its main goal is to transform a large set of variables into a smaller, more manageable set while retaining as much of the original information (variance) as possible. The new variables, or principal components, are linear combinations of the original variables and are uncorrelated.
Which of the following statements accurately describes a key difference between K-Means and DBSCAN clustering algorithms?
Answer: DBSCAN can identify noise points (outliers) that do not belong to any cluster, whereas K-Means assigns every point to a cluster.
A significant advantage of DBSCAN is its ability to handle noise and identify outliers. Points in low-density regions are marked as noise and do not get assigned to any cluster. In contrast, K-Means forces every single data point into one of the K clusters, which can distort the cluster shapes if outliers are present.
A financial institution is analyzing a large dataset of credit card transactions to identify fraudulent activities. These fraudulent transactions are rare and often exhibit unusual patterns compared to the vast majority of legitimate transactions. Which type of unsupervised learning is most appropriate for this specific task?
Answer: Anomaly Detection
Anomaly detection (or outlier detection) is the branch of unsupervised learning focused on identifying rare items, events, or observations that differ significantly from the majority of the data. Since fraudulent transactions are rare and deviate from normal patterns, this is the most suitable approach.
In agglomerative hierarchical clustering, what does the 'linkage criterion' (e.g., single, complete, average) determine?
Answer: The method for measuring the distance between clusters to decide which ones to merge next.
The linkage criterion defines how the distance between two clusters is calculated at each step of the agglomerative process. For example, 'complete' linkage uses the maximum distance between points in the two clusters, 'single' linkage uses the minimum distance, and 'average' uses the average distance. This criterion is crucial for determining the merge order and the final hierarchy.
A retail manager wants to analyze transaction data to discover which products are frequently purchased together, with the goal of optimizing store layout. For example, they want to find rules like "If a customer buys bread, they are 80% likely to also buy milk." Which unsupervised technique is best suited for this 'market basket analysis'?
Answer: Apriori Algorithm
The Apriori algorithm is a classic method for association rule mining, specifically designed to perform market basket analysis. It identifies frequent itemsets in a transactional database and derives association rules to uncover patterns of co-occurrence, which directly addresses the manager's goal.