Skip to content

Introduction to Machine Learning

SWE · Final · Fall 25

This browser can’t show the PDF here

Open it, and your phone’s PDF viewer takes it, or download it.

Read the questions as text

Read off the PDF by AI, so it may have mistakes. The PDF is the original.

Daffodil International University Department of Software Engineering Faculty of Science & Information Technology Final Examination, Fall 2025 Course Code: SE 544; Course Title: Introduction to Machine Learning Sections & Teachers: 40(B,D,G,F)(JJS), 40(E,H)(FA), 40(A,C)(NS), 40I(DB) Time: 2:00 Hours Marks: 40 Answer ALL Questions [The figures in the right margin indicate the full marks and corresponding course outcomes. All portions of each question must be answered sequentially.] 1. a) A doctor is designing a simple prediction tool that takes a patient's health score from blood sugar levels and estimates the probability that the patient has a certain disease. The model gives z=1.4 for a patient's blood sugar reading. Compute the predicted probability of diabetes using the sigmoid function. Round to 3 decimal places. Interpret what advice might the doctor give if the threshold is 0.5? [Marks-2] [CLO-1, Level-1] b) Describe the fundamental difference with examples between sigmoid function and SoftMax function. [Marks-3] [CLO-1, Level-2] 2. a) A data scientist is evaluating the performance of a binary classifier on an imbalanced medical dataset where the number of positive cases is very small compared to negative cases. Explain what a Precision-Recall (PR) curve represents and how it helps in evaluating classifier performance in such imbalanced settings. [Marks-4] [CLO-2, Level-2] b) A machine learning company is developing a binary classifier to detect whether patients have a particular disease (Positive) or not (Negative) based on medical test data. The classifier outputs probabilities for each patient. To evaluate its performance, the company computes True Positive Rate (TPR) and False Positive Rate (FPR). (FPR, TPR): (0.0, 0.0), (0.37, 0.17), (0.59, 0.35), (0.73, 0.69), (1.0, 1.0) Calculate and Plot the ROC curve using the data above clearly and interpret the Curve. Also, Calculate the Area Under the Curve (AUC) using the trapezoidal rule. Interpret the AUC value. [Marks-6] [CLO-2, Level-3] 3. a) Daffodil International University cricket team decides whether to play a match (Yes) or cancel (No) based on four weather attributes: Outlook (Sunny / Rain / Overcast), Temperature (Hot / Mild / Cool), and Humidity (High / Normal). The dataset (09 days) is: Day | Outlook | Temp. | Humidity | Decision 1 | Sunny | Hot | High | No 2 | Sunny | Hot | High | No 3 | Rain | Mild | High | No 4 | Rain | Cool | Normal | No 5 | Overcast | Cool | Normal | Yes 6 | Sunny | Mild | High | No 7 | Sunny | Cool | Normal | Yes 8 | Rain | Mild | Normal | No 9 | Sunny | Mild | Normal | Yes a. Construct a CART decision tree (use Gini impurity to choose splits). Using your tree, predict the decision and justify it for Outlook = Sunny and Humidity = Normal. b. Calculate the Gini impurity of Decision attribute. Explain what Gini Impurity is and how it helps the decision tree choose the best feature to split on. [Marks-6] [CLO-3, Level-3] b) A startup company is performing customer behavior analysis using location-based features. They collected a small sample dataset of 5 customers, each represented using two features: Annual Visits to Store (x) and Average Purchase Amount (y). The company wants to group similar customers using Agglomerative Hierarchical Clustering (Single Linkage). Five Sample Data Points are given below: C1(10,20), C2(20,10), C3(50,50), C4(60,50), C5(80,70). a. Construct the Agglomerative Hierarchical Clustering tree (dendrogram) using Euclidean distance. Show the cluster merge order. b. If you cut the dendrogram at height = 30, determine the number of clusters. Explain what height (distance) means in hierarchical intra and inter clustering. At height = 30, clusters are formed. Calculate the total inertia for clusters at that height. [Marks-6] [CLO-3, Level-3 / CLO-3, Level-2] 4. a) A health-analytics startup is building a model to classify patients into risk groups based on 20 medical indicators. Since collecting all 20 indicators is expensive, the team wants to reduce the number of input features. Describe the main types of feature selection and their sub-types and describe how each method could be applied in this medical data scenario. [Marks-5] [CLO-4, Level-1]