Data Mining-I — DU B.Sc. (Prog) Physical Science courses pertaining to Computer Science Previous Year Question Papers

S. Parts of the question must be answered together. Q1 (a) Data Mining datasets involve non-traditional types of data and data (2) distributions. What are non-traditional types of data in Data Mining? Give any two examples. (b) Why is the Naive Bayes classifier considered an early learner? Briefly (2) justify your answer. (c) Identify whether or not the given activities are data mining tasks. Justify (3) your choice. i. Extracting all records of students scoring above 90% from a database. ii. Discovering groups of customers with similar buying behavior. iii. Identifying fraudulent credit card transactions from transaction history. (d) A researcher applies Leave-One-Out Cross-Validation (LOOCV) on a (3) dataset containing n instances. Determine the number of iterations involved. What is the composition of training and test set in each iteration? (e) Is regression considered a supervised learning task or an unsupervised (3) learning task? Justify your answer. (f) In Naïve Bayes classifier, if certain combination of attributes and class (4) labels are never observed, it results in a zero conditional probability. What problems may arise in this scenario? Which alternate estimates can be used in such a scenario? (g) Consider an attribute Purchased Products with possible values (4) Convert this attribute to asymmetric binary attribute. (h) Define entropy in the context of decision tree learning. How is it used to (4) measure impurity and guide the selection of the most appropriate attribute for splitting the data? (1) A data analyst is using the DBSCAN algorithm to identify clusters in a (5) spatial dataset of customer locations. During the analysis, the analyst needs to determine which points form dense regions and which are considered outliers. Explain how the parameters Eps and Mints are used in this context, suitable reasoning. and describe how points are classified as core, border, or noise points with Q2. (a) A classifier is used to predict whether an email is Spam or Not Spam. The following results were obtained on a test dataset: Email ID Actual Class Predicted Class Not Spam Not Spam E2 Not Spam Not Spam E3 Not Spam Not Spam E4 Not Spam Spam E5 Not Spam Not Spam E6 Not Spam Not Spam E7 Spam Spam Not Spam E9 Spam Not Spam Not Spam i. Construct the confusion matrix. (2) ii. Compute accuracy, precision, recall and F1-score. (6) iii. Identify whether the data shows class imbalance. Is accuracy alone (4) sufficient to evaluate the model? Give proper reasoning (b) Differentiate between noise and outliers with one suitable example for (3) 03. Consider the following market basket transactions: TID List of Items T1 (Atta, Sugar, Tea) T2 (Sugar, Biscuits} T3 (Sugar, Ghee) (Atta, Sugar, Biscuits) T5 (Atta, Ghee) T6 (Sugar, Ghee) T7 (Atta, Gheel T8 (Atta, Sugar, Ghee, Tea) (Atta, Sugar, Ghee ) (a) Apply the Apriori algorithm on the given transactional dataset to compute (8)