DPDC — Junior Assistant Manager (JAM) — Exam: 27.06.2025 (BUET)প্রশ্ন ও সমাধান

Aspect	Supervised Learning	Unsupervised Learning
Training data	Labeled data (input + known correct output)	Unlabeled data (only inputs, no target output)
Goal	Learn a mapping from input to a known output	Discover hidden structure/patterns in the data
Feedback	Direct — error measured against true labels	None — no "correct answer" to compare against
Common tasks	Classification, Regression	Clustering, Dimensionality Reduction, Association
Examples	Spam detection (labeled spam/not-spam emails), house price prediction	Customer segmentation (grouping similar customers), anomaly detection
Example algorithms	Linear/Logistic Regression, Decision Trees, SVM, Neural Networks	K-Means Clustering, Hierarchical Clustering, PCA

graph LR
    subgraph "Supervised"
    D1["Labeled Data (X, y)"] --> M1["Train Model"] --> P1["Predict y for new X"]
    end
    subgraph "Unsupervised"
    D2["Unlabeled Data (X only)"] --> M2["Find Patterns"] --> P2["Clusters / Structure"]
    end

Example distinguishing the two: given a dataset of customer purchase histories —

• Supervised: if each record is labeled "will churn" / "will not churn," a classifier learns to predict churn for new customers.
• Unsupervised: with no labels at all, clustering (e.g., K-Means) groups customers into segments (e.g., "frequent budget buyers," "occasional premium buyers") purely from purchasing pattern similarity.

Exam takeaway: Supervised learning requires labeled training data and predicts a known target; unsupervised learning works on unlabeled data to reveal structure (clusters, associations) with no predefined "correct" output.