2026 question paper

Data Mining

26 questions

Q1a. A data cube contains 5 products, 4 regions and 12 months. The total number of base-level cells is (i) 120 (ii) 240 (iii) 180 (iv) 36020262m
Module 1: Data Warehousing and Business Analysis
A data cube contains 5 products, 4 regions and 12 months. The total number of base-level cells is

(i) 120
(ii) 240
(iii) 180
(iv) 360

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (ii) 240

Explanation

In a data cube, the number of base-level (apex-excluded, lowest-level) cells is the product of the number of distinct values in each dimension:

Cells = products x regions x months = 5 x 4 x 12 = 240

Each cell stores one measure value (for example total sales) for one unique combination of product, region and month.

Why other options are incorrect:

(i) 120 - ignores one dimension (5 x 4 x 6 or similar).

(iii) 180 and (iv) 360 - not the product of 5, 4 and 12.

Answer: (ii) 240
Q1b. Which schema generally contains a central fact table connected directly to multiple dimension tables? (i) Network schema (ii) Hierarchical schema (iii) Star schema (iv) Object schema20262m
Module 1: Data Warehousing and Business Analysis
Which schema generally contains a central fact table connected directly to multiple dimension tables?

(i) Network schema
(ii) Hierarchical schema
(iii) Star schema
(iv) Object schema

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (iii) Star schema

Explanation

A star schema has one large central fact table (measures such as sales amount, plus foreign keys) directly connected to several dimension tables (such as time, product, customer, location). The diagram looks like a star, and the dimension tables are usually denormalized, which keeps queries simple and fast.

Why other options are incorrect:

(i) Network schema and (ii) Hierarchical schema - old database models, not data warehouse schemas.

(iv) Object schema - used in object-oriented databases.

Key Points to Remember:

Snowflake schema = star schema with normalized dimension tables.

Fact constellation = several fact tables sharing dimensions.

Answer: (iii) Star schema
Q1c. If a dataset contains extreme outliers that strongly affect the mean, which measure is generally more robust for describing central tendency? (i) Mean (ii) Variance (iii) Median (iv) Standard deviation20262m
Module 2: Data Mining Fundamentals and Association Rule Mining
If a dataset contains extreme outliers that strongly affect the mean, which measure is generally more robust for describing central tendency?

(i) Mean
(ii) Variance
(iii) Median
(iv) Standard deviation

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (iii) Median

Explanation

The median is the middle value of the sorted data. It depends only on the position of values, so a few extreme outliers barely change it. The mean uses every value, so one huge value can pull it far from the typical value.

Example: data 10, 12, 13, 14, 500. Mean = 109.8 (misleading), median = 13 (representative).

Why other options are incorrect:

(i) Mean - strongly affected by outliers.

(ii) Variance and (iv) Standard deviation - these measure spread, not central tendency, and are also sensitive to outliers.

Answer: (iii) Median
Q1d. In a database of 200 transactions, an itemset occurs in 50 transactions. Its support is (i) 10% (ii) 25% (iii) 20% (iv) 50%20262m
Module 2: Data Mining Fundamentals and Association Rule Mining
In a database of 200 transactions, an itemset occurs in 50 transactions. Its support is

(i) 10%
(ii) 25%
(iii) 20%
(iv) 50%

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (ii) 25%

Explanation

Support of an itemset = (number of transactions containing the itemset) / (total number of transactions)

Support = 50 / 200 = 0.25 = 25%

Why other options are incorrect:

(i) 10% = 20/200, (iii) 20% = 40/200, (iv) 50% = 100/200.

Answer: (ii) 25%
Q1e. Support of an itemset represents (i) Probability that the consequent occurs given the antecedent (ii) Number of attributes in the database (iii) Frequency with which the itemset occurs in the database (iv) Number of rules generated20262m
Module 2: Data Mining Fundamentals and Association Rule Mining
Support of an itemset represents

(i) Probability that the consequent occurs given the antecedent
(ii) Number of attributes in the database
(iii) Frequency with which the itemset occurs in the database
(iv) Number of rules generated

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (iii) Frequency with which the itemset occurs in the database

Explanation

Support tells how often an itemset appears in the transaction database (as a count or as a fraction/percentage of all transactions). Itemsets whose support is at least the minimum support threshold are called frequent itemsets.

Why other options are incorrect:

(i) This describes confidence: P(consequent | antecedent).

(ii) Number of attributes and (iv) Number of rules - not what support means.

Key Points to Remember:

Support(A => B) = P(A and B)

Confidence(A => B) = support(A and B) / support(A)

Answer: (iii) Frequency with which the itemset occurs in the database
Q1f. A classifier correctly classifies 180 out of 200 instances. Its accuracy is (i) 80% (ii) 90% (iii) 85% (iv) 95%20262m
Module 3: Classification and Prediction
A classifier correctly classifies 180 out of 200 instances. Its accuracy is

(i) 80%
(ii) 90%
(iii) 85%
(iv) 95%

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (ii) 90%

Explanation

Accuracy = (correctly classified instances) / (total instances)

Accuracy = 180 / 200 = 0.90 = 90%

In confusion-matrix terms: Accuracy = (TP + TN) / (TP + TN + FP + FN).

Why other options are incorrect:

(i) 80% = 160/200, (iii) 85% = 170/200, (iv) 95% = 190/200.

Answer: (ii) 90%
Q1g. Which technique is particularly useful for reducing overfitting and obtaining a more reliable estimate of classifier performance? (i) Cross-validation (ii) Increasing noise (iii) Data duplication (iv) Removing the test set20262m
Module 3: Classification and Prediction
Which technique is particularly useful for reducing overfitting and obtaining a more reliable estimate of classifier performance?

(i) Cross-validation
(ii) Increasing noise
(iii) Data duplication
(iv) Removing the test set

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (i) Cross-validation

Explanation

In k-fold cross-validation the data is split into k parts (folds). The model is trained on k-1 folds and tested on the remaining fold, repeating k times so every fold is used once for testing. The average score gives a more reliable estimate of performance, and because the model is tested on unseen data each time, it helps detect and reduce overfitting.

Why other options are incorrect:

(ii) Increasing noise - makes the model worse.

(iii) Data duplication - causes test data leakage and over-optimistic results.

(iv) Removing the test set - removes any way to measure performance on unseen data.

Answer: (i) Cross-validation
Q1h. A major limitation of K-means is (i) It cannot process numerical data (ii) It always identifies outliers correctly (iii) It requires the number of clusters to be specified in advance (iv) It does not use distance measures20262m
Module 4: Cluster Analysis
A major limitation of K-means is

(i) It cannot process numerical data
(ii) It always identifies outliers correctly
(iii) It requires the number of clusters to be specified in advance
(iv) It does not use distance measures

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (iii) It requires the number of clusters to be specified in advance

Explanation

K-means needs the user to give k, the number of clusters, before it runs. A wrong k gives poor clusters, and the right k is usually unknown. Other limitations: sensitive to outliers, depends on initial centroids, finds only spherical clusters, works on numeric data.

Why other options are incorrect:

(i) K-means is designed for numerical data.

(ii) It does not handle outliers well; they pull centroids away.

(iv) It relies on a distance measure (usually Euclidean).

Answer: (iii) It requires the number of clusters to be specified in advance
Q1i. If a word occurs 4 times in a document containing 100 words, its term frequency is (i) 0.004 (ii) 0.4 (iii) 0.04 (iv) 420262m
Module 5: Mining Object, Spatial, Multimedia, Text and Web Data
If a word occurs 4 times in a document containing 100 words, its term frequency is

(i) 0.004
(ii) 0.4
(iii) 0.04
(iv) 4

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (iii) 0.04

Explanation

Term frequency (TF) = (number of times the term occurs in the document) / (total number of words in the document)

TF = 4 / 100 = 0.04

TF is the first part of the TF-IDF weight used in text mining: TF-IDF = TF x log(N / df).

Why other options are incorrect:

(i) 0.004 = 4/1000, (ii) 0.4 = 4/10, (iv) 4 is the raw count, not the frequency ratio.

Answer: (iii) 0.04
Q1j. Which technique is most appropriate for discovering hyperlinks and relationships among Web pages? (i) Web structure mining (ii) Numerical integration (iii) Web usage mining (iv) Data normalization20262m
Module 5: Mining Object, Spatial, Multimedia, Text and Web Data
Which technique is most appropriate for discovering hyperlinks and relationships among Web pages?

(i) Web structure mining
(ii) Numerical integration
(iii) Web usage mining
(iv) Data normalization

View this question on its own page →
Search Google Ask AI
Worked Solution
Solution

Correct Answer: (i) Web structure mining

Explanation

Web structure mining analyses the link structure of the Web: hyperlinks between pages and the relationships among them. It is used to find authoritative pages, hubs and communities. Well-known algorithms: PageRank and HITS.

The three types of web mining:

Web content mining - the content of pages (text, images)

Web structure mining - hyperlinks between pages

Web usage mining - user access logs and click patterns

Why other options are incorrect:

(ii) Numerical integration - a mathematical technique, unrelated.

(iii) Web usage mining - studies user behavior from logs, not links.

(iv) Data normalization - a preprocessing step.

Answer: (i) Web structure mining
Budget App•POCKETLY
Pocketly on Android
Log daily food, travel, and hostel expenses instantly. Set monthly spending limits and see where your money goes without complex spreadsheets.
Install Pocketly

Q2a. Explain the major components of a data warehouse and discuss the functions of data extraction, cleaning, transformation, metadata management and reporting in the data-warehouse environment.20267m

Module 1: Data Warehousing and Business Analysis

Explain the major components of a data warehouse and discuss the functions of data extraction, cleaning, transformation, metadata management and reporting in the data-warehouse environment.

View this question on its own page →

Search Google Ask AI

Worked Solution

Components of a Data Warehouse and Their Functions

A data warehouse is a subject-oriented, integrated, time-variant and non-volatile collection of data that supports management's decision-making.

Architecture

Operational DBs, files,    ETL / Staging area            Warehouse layer          Front-end tools
external sources           (extract, clean,              (DW database,            (query, reports,
                            transform, load)              data marts)              OLAP, mining)
   [Source 1] \                                                                  
   [Source 2] ---> [ Extraction -> Cleaning -> Transformation -> Load ] ---> [ Data Warehouse ] ---> [ Users ]
   [Source 3] /                         ^                                   ^
                                        |______ [ Metadata repository ] ____|

Major components

Component	Role
Data sources	Operational databases, flat files, external data feeding the warehouse
ETL / staging area	Extracts, cleans, transforms and loads data
Data warehouse database	Central integrated store, usually on a relational DBMS or multidimensional server
Data marts	Subject-specific subsets (sales, finance)
Metadata repository	Data about the data: schema, mappings, rules, lineage
Query, reporting and OLAP tools	Let users analyse and view the data
Management tools	Scheduling, security, backup, monitoring

Functions in the warehouse environment

Data extraction: Pulls the required data from heterogeneous sources, either fully or incrementally (only changes), without disturbing operational systems.
Data cleaning: Detects and fixes errors: missing values, duplicates, wrong formats, inconsistent codes. This makes the warehouse data reliable.
Data transformation: Converts data into one consistent format: unit and format conversion, key mapping, aggregation, derived fields, and applying business rules.
Metadata management: Stores and maintains technical metadata (schema, mappings, ETL logic), business metadata (definitions, ownership) and operational metadata (load times, history). It helps users find and trust data and supports lineage.
Reporting: Gives users standard and ad hoc reports, dashboards, OLAP analysis and data mining outputs that turn warehouse data into decisions.

Conclusion

ETL fills the warehouse with clean, consistent data, metadata describes and controls it, and query/reporting tools turn it into information for decision-making.

Q2b. A data warehouse contains sales data for 4 regions, 12 months and 5 product categories. Each combination stores the total sales value. Determine: (i) The total number of possible multidimensional cells, (ii) The number of cells required for one region across all months and products and (iii) Explain how roll-up, drill-down, slice and dice operations can be performed on this cube.20267m

Module 1: Data Warehousing and Business Analysis

A data warehouse contains sales data for 4 regions, 12 months and 5 product categories. Each combination stores the total sales value. Determine:

(i) The total number of possible multidimensional cells,
(ii) The number of cells required for one region across all months and products and
(iii) Explain how roll-up, drill-down, slice and dice operations can be performed on this cube.

View this question on its own page →

Search Google Ask AI

Worked Solution

Data Cube: Sales by Region, Month and Product Category

Dimensions: Region = 4, Month = 12, Product category = 5. Measure: total sales.

(i) Total number of possible multidimensional cells

Each combination of (region, month, category) is one cell:

Cells = 4 x 12 x 5 = 240

(ii) Cells required for one region across all months and products

Fix the region (1 value) and vary the other two dimensions:

Cells = 1 x 12 x 5 = 60

(iii) OLAP operations on this cube

Operation	What it does	Example on this cube	Result
Roll-up	Aggregate to a higher level of a dimension hierarchy (or remove a dimension)	Months to quarters (12 to 4); sales then shown per region, quarter and category	4 x 4 x 5 = 80 cells
Drill-down	Go from summary to more detailed data	Quarter to month, or month to day; Region to city	More cells than before (e.g. back to 240)
Slice	Select one value on one dimension, giving a sub-cube with one dimension fewer	Region = "North": a 2-D table of month x category	12 x 5 = 60 cells
Dice	Select ranges/values on two or more dimensions, giving a smaller cube	Regions {North, South}, months Jan to Mar, categories {Electronics, Clothing}	2 x 3 x 2 = 12 cells
Pivot (rotate)	Rotate axes to see the data from another view	Swap month and category axes	Same cells, new view

Extra: number of cuboids

With 3 dimensions and no hierarchies, the cube has 2^3 = 8 cuboids: the base cuboid (region, month, category), three 2-D cuboids, three 1-D cuboids and the apex (grand total).

Conclusion

The cube holds 240 base cells; one region slice needs 60 cells; roll-up, drill-down, slice and dice let a user move between summary and detail views.

Q3a. A dataset contains numerical, categorical and missing attributes. Develop a suitable preprocessing pipeline involving data cleaning, integration, transformation, reduction and discretization. Justify the choice of each operation.20267m

Module 2: Data Mining Fundamentals and Association Rule Mining

A dataset contains numerical, categorical and missing attributes. Develop a suitable preprocessing pipeline involving data cleaning, integration, transformation, reduction and discretization. Justify the choice of each operation.

View this question on its own page →

Search Google Ask AI

Worked Solution

Preprocessing Pipeline for a Dataset with Numerical, Categorical and Missing Attributes

Raw data is usually incomplete, noisy and inconsistent. The steps below are applied in this order: clean, integrate, transform, reduce, discretize.

Pipeline

Step	What to do	Justification
1. Data cleaning	Missing values: numerical attributes get mean/median (median if skewed); categorical attributes get the mode or an "Unknown" label; if over about 50% of a column is missing, drop it. Noise: smooth by binning or regression. Outliers: detect with IQR or z-score; cap or remove if they are errors. Inconsistency: standardize codes and formats (e.g. "M"/"Male")	Algorithms cannot handle missing values and noise lowers accuracy
2. Data integration	Merge multiple sources, resolve schema and entity conflicts (same customer, different IDs), remove duplicates. Check redundant attributes with correlation (numerical) or chi-square (categorical)	Avoids duplicates and conflicting values and keeps one consistent view
3. Data transformation	Normalize numerical attributes (min-max to [0, 1] or z-score); encode categorical attributes (one-hot, label encoding); apply aggregation or log transform for skewed data	Distance-based methods (KNN, K-means) would otherwise be dominated by large-scale attributes; most algorithms need numbers
4. Data reduction	Attribute subset selection (drop irrelevant/redundant), PCA for dimensionality reduction, sampling (numerosity reduction) if the data is huge	Lowers storage and computation time and reduces the curse of dimensionality
5. Discretization	Convert continuous attributes into intervals using equal-width or equal-frequency binning, and build concept hierarchies (age to young / middle / old)	Needed for association rules and decision trees, and gives more understandable patterns

Why this order

Cleaning must come first, because wrong or missing values would distort later steps. Integration comes before transformation so that scaling is done on the final merged data. Reduction and discretization come last, once the data is clean and consistent.

Example

Attribute Income (numerical, 8% missing): fill with median, cap outliers at the 99th percentile, normalize to [0, 1], then bin into Low / Medium / High. Attribute City (categorical): fix spelling variants and one-hot encode.

Conclusion

The pipeline produces complete, consistent, scaled and compact data, which gives more accurate and reliable mining results.

Q3b. Explain the major functionalities of data mining. Discuss how data preprocessing improves the quality and reliability of knowledge discovered from large datasets.20267m

Module 2: Data Mining Fundamentals and Association Rule Mining

Explain the major functionalities of data mining. Discuss how data preprocessing improves the quality and reliability of knowledge discovered from large datasets.

View this question on its own page →

Search Google Ask AI

Worked Solution

Functionalities of Data Mining and the Role of Preprocessing

Data mining is the process of discovering interesting patterns and knowledge from large amounts of data. Its functionalities specify the kinds of patterns that can be found.

Major functionalities

Functionality	Description	Example
Concept/class description (characterization and discrimination)	Summarizes the general features of a class, or compares classes	Features of customers who buy a lot vs those who buy little
Association and correlation analysis	Finds items that occur together (frequent patterns)	Bread and butter bought together (market basket)
Classification	Builds a model from labelled data to predict a class label (supervised)	Classify loan applicants as safe or risky
Prediction / regression	Predicts a continuous value	Forecast next month's sales
Clustering	Groups similar objects with no predefined labels (unsupervised)	Customer segmentation
Outlier analysis	Finds objects that do not fit the general behaviour	Credit card fraud detection
Evolution / trend analysis	Studies how objects change over time	Stock price trends

How preprocessing improves quality and reliability

Real-world data is often incomplete, noisy and inconsistent; mining such data gives unreliable results ("garbage in, garbage out"). Preprocessing improves the data on these quality measures:

Quality factor	How preprocessing helps
Accuracy	Cleaning removes noise and errors
Completeness	Missing values are filled or handled
Consistency	Integration and standardization remove conflicts and duplicates
Believability	Verified, cleaned data is trusted more
Interpretability	Discretization and concept hierarchies make patterns easier to understand
Efficiency	Reduction cuts data size and mining time

Effects on the discovered knowledge

Cleaning prevents wrong patterns caused by errors and outliers.
Integration gives a single, consistent view across sources.
Transformation (normalization) stops large-scale attributes from dominating distance-based algorithms.
Reduction removes irrelevant attributes, so models are simpler and less prone to overfitting.
Discretization makes rules concise and suitable for algorithms like decision trees and Apriori.

Conclusion

Good preprocessing improves the accuracy, stability and trustworthiness of classification, clustering and association results, and often takes most of the effort in a data mining project.

Q4a. Explain the Apriori principle and analyze how candidate generation and pruning reduce the search space during frequent-itemset mining. Compare Apriori with a scalable frequent-pattern mining approach such as FP-Growth.20267m

Module 2: Data Mining Fundamentals and Association Rule Mining

Explain the Apriori principle and analyze how candidate generation and pruning reduce the search space during frequent-itemset mining. Compare Apriori with a scalable frequent-pattern mining approach such as FP-Growth.

View this question on its own page →

Search Google Ask AI

Worked Solution

Apriori Principle, Candidate Generation and Pruning, and Comparison with FP-Growth

Apriori principle

All non-empty subsets of a frequent itemset must also be frequent. Equivalently, if an itemset is infrequent, every superset of it is also infrequent (the anti-monotone property of support).

Example: if {A, D} is infrequent, then {A, B, D}, {A, C, D} and {A, B, C, D} cannot be frequent, so they need not be checked.

How Apriori works (level-wise search)

Scan the database and find frequent 1-itemsets L1.
Candidate generation (join step): join Lk-1 with itself to produce candidate k-itemsets Ck.
Pruning: remove any candidate that has a (k-1)-subset that is not in Lk-1.
Scan the database to count the support of the remaining candidates; keep those with support at least min_sup as Lk.
Repeat until no new frequent itemsets are found.

How candidate generation and pruning reduce the search space

With n items there are up to 2^n - 1 possible itemsets, too many to test.

Join only combines frequent itemsets, so only itemsets that could be frequent are generated.
Pruning with the Apriori property removes candidates before the costly database scan.

Example: L2 = {AB, AC, AD, BC}. The join step (itemsets sharing the first item) gives ABC, ABD and ACD. ABC survives, because AB, AC and BC are all frequent. ABD is pruned because BD is not in L2, and ACD is pruned because CD is not in L2. Only ABC is counted in the database scan.

Apriori vs FP-Growth

Basis	Apriori	FP-Growth
Approach	Generate-and-test (candidate generation)	Pattern growth using an FP-tree
Candidate generation	Required; can be huge	None
Database scans	One per level (many scans)	Two scans in total
Data structure	Candidate sets and hash trees	Compact FP-tree plus header table
Strategy	Breadth-first, level-wise	Divide-and-conquer (conditional pattern bases)
Speed on large or dense data	Slow	Much faster
Memory	Large candidate sets	FP-tree must fit in memory; compact for dense data
Complexity to implement	Simple	More complex

Conclusion

Apriori is simple and prunes well using the anti-monotone property, but it is costly because of repeated scans and many candidates. FP-Growth removes candidate generation by compressing the data into an FP-tree, so it is more scalable.

Q4b. Explain different types of association rules, including Boolean, quantitative, single-dimensional, multidimensional, positive and negative association rules. Discuss situations in which each type is useful.20267m

Module 2: Data Mining Fundamentals and Association Rule Mining

Explain different types of association rules, including Boolean, quantitative, single-dimensional, multidimensional, positive and negative association rules. Discuss situations in which each type is useful.

View this question on its own page →

Search Google Ask AI

Worked Solution

Types of Association Rules and When Each Is Useful

An association rule A => B states that when A occurs, B tends to occur too, measured by support and confidence.

Types of association rules

Type	Meaning	Example	Useful when
Boolean	Items are only present or absent	computer => antivirus software	Market basket analysis with yes/no purchase data
Quantitative	Involves numeric attributes, which are discretized into ranges	age(30-39) AND income(42K-48K) => buys(HDTV)	Data has numeric attributes such as age, salary, quantity
Single-dimensional	Only one predicate (dimension) is used	buys(X, "milk") => buys(X, "bread")	Simple item-to-item patterns in one table of transactions
Multidimensional	Two or more dimensions or predicates	age(X, "20-29") AND occupation(X, "student") => buys(X, "laptop")	Data warehouses and customer data with many attributes
Positive	Rule on presence of items (A => B)	tea => sugar	Finding items to bundle, promote or place together
Negative	Rule involving absence: A => NOT B, or NOT A => B	buys tea => NOT buys coffee	Finding substitutes and competing products, fraud and anomaly cases

Related types (for completeness)

Multilevel rules: use concept hierarchies ("milk" => "bread" vs "skim milk" => "wheat bread").
Single-level rules: all items at the same abstraction level.

Situations where each helps

Boolean / single-dimensional / positive: supermarket layout, cross-selling and recommendations.
Quantitative: targeting customers by age or income bands, pricing and credit analysis.
Multidimensional: customer segmentation and campaigns using several attributes together.
Negative: detecting substitute products (so you do not stock both aggressively) and unusual behaviour.

Evaluating the rules

Strong rules meet minimum support and confidence; lift is used to check that A and B are really related (lift > 1 means positive correlation).

Conclusion

The choice of rule type depends on the data (binary or numeric, one or many attributes) and the business goal.

Q5. Design an experimental framework for evaluating a classification model. Explain the use of training/testing datasets, cross-validation, confusion matrix, accuracy, precision, recall, F1-score and ROC-based evaluation. Discuss how ensemble methods can improve predictive performance.202614m

Module 3: Classification and Prediction

Design an experimental framework for evaluating a classification model. Explain the use of training/testing datasets, cross-validation, confusion matrix, accuracy, precision, recall, F1-score and ROC-based evaluation. Discuss how ensemble methods can improve predictive performance.

View this question on its own page →

Search Google Ask AI

Worked Solution

Experimental Framework for Evaluating a Classification Model

The aim is to measure how well the model will work on unseen data, reliably and fairly.

Step 1: Prepare the data

Clean the data, handle missing values, encode and scale features. Fix a random seed so results can be repeated.

Step 2: Training and testing datasets

Hold-out: split the data, for example 70% training / 30% testing (or 80/20). Use a stratified split so the class proportions are the same in both sets.
Training set: used to build (fit) the model.
Validation set (optional): used to tune hyperparameters (e.g. tree depth).
Test set: used only once at the end; never used for training or tuning, to avoid leakage.

Step 3: Cross-validation

k-fold cross-validation (k = 10 is common):

Split the data into k equal folds.
Train on k-1 folds, test on the remaining fold.
Repeat k times so every fold is used once for testing.
Average the k scores.

This gives a more reliable estimate than a single split and reduces overfitting. Variants: stratified k-fold, leave-one-out (small data) and bootstrap.

Step 4: Confusion matrix

	Predicted Positive	Predicted Negative
Actual Positive	TP	FN
Actual Negative	FP	TN

Step 5: Evaluation measures

Measure	Formula	Meaning
Accuracy	(TP + TN) / (TP + TN + FP + FN)	Overall fraction correct; misleading for imbalanced classes
Precision	TP / (TP + FP)	Of the predicted positives, how many are truly positive
Recall (sensitivity, TPR)	TP / (TP + FN)	Of the actual positives, how many were found
F1-score	2 x Precision x Recall / (Precision + Recall)	Harmonic mean of precision and recall
Specificity	TN / (TN + FP)	Of the actual negatives, how many were correct
False positive rate (FPR)	FP / (FP + TN)	Used on the ROC curve

Example: TP = 40, FP = 10, FN = 5, TN = 45 (100 instances)

Accuracy = 85/100 = 0.85
Precision = 40/50 = 0.80
Recall = 40/45 = 0.889
F1 = 2(0.80)(0.889)/(0.80 + 0.889) = 0.842

Step 6: ROC-based evaluation

The ROC curve plots TPR (y-axis) against FPR (x-axis) at different classification thresholds.

A curve close to the top-left corner means a good model.
The diagonal line is random guessing.
AUC (area under the curve) summarizes performance: 1.0 is perfect, 0.5 is random.
ROC is useful for comparing models and for imbalanced data, together with precision-recall curves.

Step 7: Compare and report

Compare against a baseline (such as majority class), compare several models on the same folds, report the mean and standard deviation of the metrics, and check with a significance test if needed.

How ensemble methods improve predictive performance

An ensemble combines several models so that the combined prediction is better than any single model.

Method	Idea	Effect
Bagging (e.g. Random Forest)	Train models on bootstrap samples; vote or average	Reduces variance and overfitting
Boosting (e.g. AdaBoost, Gradient Boosting)	Train models in sequence; each focuses on the previous errors	Reduces bias, gives high accuracy
Stacking	A meta-model learns to combine the outputs of different models	Uses the strengths of diverse models

Ensembles work because different models make different errors, and combining them cancels many of those errors. Ensembles should themselves be evaluated with the same cross-validation framework.

Conclusion

A sound framework uses a clean train/test split, cross-validation, several metrics (accuracy, precision, recall, F1, ROC-AUC) and, where needed, ensembles, to get a trustworthy estimate of how the model will perform in practice.

Q6. Design a clustering framework for identifying customer segments using a large multidimensional dataset. Discuss preprocessing, feature selection, choice of clustering algorithm, selection of the number of clusters, outlier handling and cluster validation.202614m

Module 4: Cluster Analysis

Design a clustering framework for identifying customer segments using a large multidimensional dataset. Discuss preprocessing, feature selection, choice of clustering algorithm, selection of the number of clusters, outlier handling and cluster validation.

View this question on its own page →

Search Google Ask AI

Worked Solution

Clustering Framework for Customer Segmentation

Goal: group customers with similar behaviour so that marketing, pricing and service can be targeted per segment. Clustering is unsupervised: there are no predefined labels.

Framework overview

Business goal -> Data collection -> Preprocessing -> Feature selection
   -> Choose algorithm -> Choose number of clusters -> Handle outliers
   -> Validate clusters -> Profile segments and act

1. Data and business goal

Define what "similar" means. Typical features: RFM (recency, frequency, monetary value), age, income, location, product categories, web and app activity, tenure.

2. Preprocessing

Missing values: impute (median/mode) or drop sparse columns.
Duplicates and inconsistencies: merge and standardize.
Categorical features: one-hot or label encoding (or use a distance made for mixed data).
Scaling: min-max or z-score, so large-range attributes (income) do not dominate distance.
Skewed features: log-transform spending data.

3. Feature selection and dimensionality reduction

In high-dimensional data, distances become less meaningful (curse of dimensionality).

Remove irrelevant and highly correlated features (correlation filter, variance threshold).
Use PCA to compress correlated features into fewer components.
Keep features that are meaningful to the business.

4. Choice of clustering algorithm

Algorithm	Best when	Limitation
K-means	Large data, roughly spherical clusters, numeric features	Needs k; sensitive to outliers
Hierarchical	Smaller data, want a dendrogram	Costly for large data
DBSCAN	Arbitrary shapes, noise present, k unknown	Struggles with varying densities
Gaussian mixture (model-based)	Overlapping segments, soft membership	Assumes distribution form
K-prototypes / K-medoids	Mixed numeric and categorical data	Slower

For a large customer dataset, K-means (or mini-batch K-means) is a good start; DBSCAN or GMM can be used to compare.

5. Selecting the number of clusters

Elbow method: plot within-cluster sum of squares (WCSS) against k and pick the bend.
Silhouette score: choose the k with the highest average silhouette.
Gap statistic, or dendrogram cut for hierarchical clustering.
Business usefulness: the segments must be few enough to act on (often 3 to 8).

6. Outlier handling

Outliers pull centroids and distort K-means.

Detect with IQR, z-score, Local Outlier Factor or DBSCAN noise points.
Decide: remove or cap them if they are errors; keep them as a separate group if they are genuine (very high-value customers or fraud).

7. Cluster validation

Type	Measure
Internal (no labels)	Silhouette coefficient (higher is better), Davies-Bouldin index (lower is better), Calinski-Harabasz index (higher is better)
External (if labels exist)	Adjusted Rand Index, purity
Stability	Re-run with different seeds or samples; clusters should stay similar
Business validation	Segments are distinct, interpretable and actionable

8. Interpretation and use

Profile each cluster by its mean feature values, for example high-value loyal, new customers, discount seekers, at risk of churn, and design offers for each. Monitor and re-cluster periodically as behaviour changes.

Conclusion

A good segmentation framework combines careful preprocessing, sensible feature selection, a suitable algorithm, a justified k, robust outlier handling and validation, ending in segments the business can act on.

Q7a. Explain the challenges associated with mining spatial, multimedia, text and Web data. Discuss why conventional data-mining techniques may require modification for complex data objects.20267m

Module 5: Mining Object, Spatial, Multimedia, Text and Web Data

Explain the challenges associated with mining spatial, multimedia, text and Web data. Discuss why conventional data-mining techniques may require modification for complex data objects.

View this question on its own page →

Search Google Ask AI

Worked Solution

Challenges in Mining Spatial, Multimedia, Text and Web Data

These are complex data objects: they have rich structure, relationships between objects, or no fixed format, unlike simple records in a table.

Challenges by data type

Data type	Main challenges
Spatial (maps, GIS, satellite data)	Objects are related by location and distance, so records are not independent (spatial autocorrelation); complex data types (points, lines, polygons); topological relations (near, inside, overlaps); different scales and resolutions; very large volume; uncertain or noisy measurements
Multimedia (images, audio, video)	Huge size; high-dimensional features; the semantic gap (low-level features such as colour or pitch vs human meaning); similarity search is costly; temporal order in video and audio; lack of labels
Text	Unstructured; very high dimensionality and sparse vectors; ambiguity (synonyms, polysemy); language dependence, slang, spelling errors; context and sarcasm are hard to capture
Web	Enormous and growing size; dynamic, constantly changing pages; heterogeneous formats; noisy, low-quality and spam content; need to use hyperlink structure and user logs; privacy concerns; distributed data

Why conventional data mining techniques need modification

Classic algorithms (Apriori, decision trees, K-means) assume flat, structured records with independent objects and simple numeric or categorical attributes. Complex objects break these assumptions:

Dependence between objects: nearby spatial objects or linked web pages influence each other, so the independence assumption fails; methods must use spatial or link relationships.
No ready attributes: images, audio and text must first go through feature extraction (colour histograms, TF-IDF, embeddings) before they can be mined.
Special similarity measures: Euclidean distance on raw pixels or words is not meaningful; measures such as cosine similarity, spatial distance or edit distance are needed.
High dimensionality and sparsity: requires dimensionality reduction (PCA, LSA) and feature selection.
Scalability and indexing: need special index structures (R-trees, inverted indexes) and efficient algorithms.
Semantics and context: need ontologies, language models and domain knowledge to interpret meaning.
Mixed data: objects can combine several media (text + image + links), so integrated methods are needed.

Examples of adapted techniques

Spatial: spatial association rules, spatial clustering (DBSCAN), spatial data cubes.
Multimedia: content-based image retrieval, feature-based classification.
Text: vector space model, TF-IDF, topic models.
Web: PageRank, HITS, web usage mining on logs.

Conclusion

Because complex objects are structured, interdependent, high-dimensional and semantic, standard mining methods must be extended with feature extraction, special similarity measures, relationship-aware algorithms and scalable indexing.

Q7b. Explain the major techniques used in text mining, including document representation, term weighting, text classification, clustering and information extraction. Analyze the challenges caused by high dimensionality and semantic ambiguity.20267m

Module 5: Mining Object, Spatial, Multimedia, Text and Web Data

Explain the major techniques used in text mining, including document representation, term weighting, text classification, clustering and information extraction. Analyze the challenges caused by high dimensionality and semantic ambiguity.

View this question on its own page →

Search Google Ask AI

Worked Solution

Major Techniques in Text Mining

Text mining extracts useful patterns and knowledge from unstructured text such as emails, reviews, news and documents.

Text preprocessing (first step)

Tokenization, lower-casing, stop-word removal, stemming/lemmatization (running to run), and spelling correction.

1. Document representation

Bag of words: each document is a vector of word counts, ignoring order.
Vector space model (VSM): documents and queries are vectors in term space; similarity is measured by cosine similarity.
N-grams: sequences of n words capture some word order.
Embeddings (word2vec, BERT): dense vectors that capture meaning.

2. Term weighting

Gives more importance to informative words.

TF = (count of term in document) / (total words in document). Example: 4 occurrences in 100 words gives TF = 0.04.
IDF = log(N / df), where N is the number of documents and df the number containing the term.
TF-IDF = TF x IDF: high for a word frequent in one document but rare overall.

3. Text classification

Assign documents to predefined categories (supervised): spam filtering, sentiment analysis, topic labelling.

Methods: Naive Bayes, SVM, KNN, decision trees, neural networks.

4. Text clustering

Group similar documents without labels (unsupervised): grouping news stories or support tickets.

Methods: K-means on TF-IDF vectors, hierarchical clustering, topic models (LDA).

5. Information extraction

Pull structured facts out of text.

Named entity recognition (people, places, organizations), relation extraction (who works for whom), event extraction; filling templates and databases.

Challenges

Challenge	Problem	Remedies
High dimensionality	Vocabulary can be tens of thousands of terms; vectors are very sparse; slow and prone to overfitting (curse of dimensionality)	Stop-word removal, stemming, feature selection (chi-square, information gain), dimensionality reduction (LSA/SVD, PCA), embeddings
Semantic ambiguity	Polysemy (one word, many meanings, e.g. "bank"), synonymy (many words, one meaning), context, idioms, sarcasm, negation	Word sense disambiguation, ontologies and WordNet, latent semantic analysis, contextual embeddings, part-of-speech tagging

Other issues: multiple languages, noisy and informal text, very large corpora.

Conclusion

Text mining converts raw text into numeric representations (VSM with TF-IDF), then applies classification, clustering and extraction. Its main difficulties are the huge sparse feature space and the ambiguity of natural language, which are handled with dimensionality reduction and semantic methods.

Budget App•POCKETLY
Pocketly on Android
Log daily food, travel, and hostel expenses instantly. Set monthly spending limits and see where your money goes without complex spreadsheets.
Install Pocketly

Q8a. Compare data warehouse architectures such as single-tier, two-tier and three-tier architectures. Analyze the role of a DBMS, ETL tools, metadata repository and query/reporting tools in supporting decision-making.20267m

Module 1: Data Warehousing and Business Analysis

Compare data warehouse architectures such as single-tier, two-tier and three-tier architectures. Analyze the role of a DBMS, ETL tools, metadata repository and query/reporting tools in supporting decision-making.

View this question on its own page →

Search Google Ask AI

Worked Solution

Data Warehouse Architectures and Roles of Its Components

Comparison of single-tier, two-tier and three-tier architectures

Basis	Single-tier	Two-tier	Three-tier
Structure	One layer: the warehouse and the operational data are kept together; users query it directly	Two layers: a data source/warehouse (server) layer and a client layer, usually with data marts and a staging area	Three layers: bottom (warehouse DB server + ETL), middle (OLAP server), top (front-end tools)
Aim	Minimize the amount of stored data (remove redundancy)	Separate analysis from operational systems	Best separation and scalability
Performance	Poor: analytical queries slow down operational work	Better, but limited as the number of users grows	Best: OLAP server handles heavy analysis
Scalability	Very low	Limited (connectivity and network bottlenecks)	High
Complexity and cost	Lowest	Moderate	Highest
Use today	Rare, not suitable for real warehousing	Small or departmental warehouses	Most common in enterprises

Three-tier architecture in more detail

Top tier:     Front-end tools   (query, reports, dashboards, OLAP clients, data mining)
                    |
Middle tier:  OLAP server       (ROLAP / MOLAP / HOLAP)
                    |
Bottom tier:  Data warehouse DB  <--  ETL  <--  Operational sources + Metadata repository

Roles of components in decision-making

Component	Role
DBMS	Stores and manages warehouse data; provides indexing, partitioning, storage, security and efficient query processing for large volumes
ETL tools	Extract data from sources, clean and transform it, and load it into the warehouse, giving decision-makers accurate, consistent and integrated data
Metadata repository	Holds data about data: schema, mappings, lineage, business definitions, load history. It helps users understand and trust the data and helps administrators maintain it
Query and reporting tools	Let users run queries, produce reports and dashboards, perform OLAP analysis and data mining, so information reaches managers in usable form

How they support decisions

ETL supplies clean and consistent data; the DBMS stores it for fast access; metadata explains it; and query/reporting/OLAP tools let managers see trends, compare periods and drill into causes. Together they turn operational data into decision-ready information.

Conclusion

Single-tier is simple but impractical, two-tier is adequate for small systems, and three-tier is the preferred architecture for scalable, high-performance decision support.

Q8b. Analyze the problems associated with missing values, noisy data, inconsistent data, redundant attributes and outliers. Compare different data-cleaning strategies for handling these problems.20267m

Module 2: Data Mining Fundamentals and Association Rule Mining

Analyze the problems associated with missing values, noisy data, inconsistent data, redundant attributes and outliers. Compare different data-cleaning strategies for handling these problems.

View this question on its own page →

Search Google Ask AI

Worked Solution

Data Quality Problems and Data-Cleaning Strategies

Real data is often incomplete, noisy and inconsistent. Cleaning fixes these problems before mining.

Problems and strategies

Problem	Cause	Cleaning strategies
Missing values	Faulty entry, equipment failure, non-response	(1) Ignore the tuple (when many values are missing); (2) fill manually; (3) use a global constant ("Unknown"); (4) fill with the mean/median of the attribute; (5) use the class-wise mean; (6) use the most probable value by regression, decision tree or Bayes
Noisy data (random error or variance)	Faulty instruments, entry errors	Binning (smooth by bin means, medians or boundaries); regression (fit a function); clustering (values outside clusters are noise); human inspection
Inconsistent data	Different formats or codes across sources, violated rules	Apply integrity constraints and rules, standardize formats and codes, use reference data, correct manually or with tools, and track lineage
Redundant attributes	Same information in several attributes (derived or duplicated)	Correlation analysis (Pearson r for numeric, chi-square for categorical); drop one of the pair; attribute selection
Outliers	Errors or genuine rare events	Detect with IQR, z-score, clustering, DBSCAN, LOF; remove, cap (winsorize), transform, or keep if they are genuine and important

Comparison of the strategies

Strategy	Advantage	Disadvantage
Ignore tuple	Simple	Loses data; biased if missing is not random
Mean / median fill	Easy, keeps all records	Reduces variance; may distort relations
Most probable value (model-based)	Most accurate, uses other attributes	More complex and costly
Binning	Simple smoothing	Loses detail; choice of bins matters
Regression smoothing	Preserves the trend	Needs a suitable model
Clustering for noise/outliers	Detects groups and unusual points	Depends on the algorithm and its parameters
Correlation-based removal	Reduces redundancy and speeds up mining	Correlation does not always mean redundancy
Removing outliers	Cleaner model	May discard valuable rare events (e.g. fraud)

Choosing a strategy

Use the median for skewed data and model-based filling when accuracy matters.
Remove outliers only if they are errors; keep genuine ones, because they may be the target of the analysis (fraud detection).
Always apply cleaning rules consistently and document them.

Conclusion

Each data problem needs a suitable strategy; simple methods are fast but approximate, while model-based methods are more accurate but more expensive. Good cleaning improves the reliability of the mined results.

Q9a. Write a short note on ETL Framework. _(Q.9: short notes on any two of the following)_20267m

Module 1: Data Warehousing and Business Analysis

Write a short note on ETL Framework.

(Q.9: short notes on any two of the following)

View this question on its own page →

Search Google Ask AI

Worked Solution

ETL Framework

ETL stands for Extract, Transform, Load. It is the process that moves data from many source systems into the data warehouse in a clean, consistent form.

Flow

Sources (DBs, files, apps) -> EXTRACT -> Staging area -> TRANSFORM -> LOAD -> Data warehouse

Stages

Stage	What happens
Extract	Read the required data from heterogeneous sources (databases, flat files, APIs). Can be full extraction or incremental (only new or changed data, using timestamps or change data capture), without affecting the operational systems
Transform	Done in a staging area. Includes cleaning (fix errors, missing values, duplicates), standardization (formats, units, codes), integration (merge sources, key mapping), derivation (calculated fields), aggregation and applying business rules
Load	Write the transformed data into the warehouse: initial full load first, then periodic incremental/refresh loads; also update indexes, aggregates and data marts

Supporting elements

Staging area: temporary storage where raw data is held and transformed.
Metadata: records mappings, rules and lineage.
Scheduling, logging and error handling: to run loads on time and track rejected records.

ETL vs ELT

In ELT the raw data is loaded first and transformed inside the target system (common with cloud and big data platforms).

Tools

Informatica, Talend, Microsoft SSIS, Apache NiFi, Pentaho.

Importance and challenges

ETL gives decision-makers accurate, integrated and consistent data.
Challenges: data quality, large volumes, load-time windows, changing source schemas and keeping history.

Q9b. Write a short note on Data Preprocessing Techniques. _(Q.9: short notes on any two of the following)_20267m

Module 2: Data Mining Fundamentals and Association Rule Mining

Write a short note on Data Preprocessing Techniques.

(Q.9: short notes on any two of the following)

View this question on its own page →

Search Google Ask AI

Worked Solution

Data Preprocessing Techniques

Data preprocessing prepares raw data for mining by making it complete, consistent, accurate and compact. Real data is usually dirty (missing, noisy, inconsistent), so preprocessing is needed to get reliable results.

Main techniques

Technique	Purpose	Methods
Data cleaning	Fix errors in the data	Fill missing values (ignore tuple, mean/median, most probable value); smooth noise (binning, regression, clustering); detect outliers; resolve inconsistencies
Data integration	Combine data from several sources	Schema integration, entity resolution, remove duplicates, detect redundancy with correlation and chi-square
Data transformation	Convert data into suitable forms	Normalization, aggregation, generalization, attribute construction, encoding
Data reduction	Get a smaller representation with the same analytical results	Data cube aggregation, attribute subset selection, dimensionality reduction (PCA), numerosity reduction (sampling, histograms), compression
Data discretization	Replace numeric values with intervals or labels	Equal-width and equal-frequency binning, histogram and clustering analysis, concept hierarchies

Normalization formulas

Min-max: v' = (v - min) / (max - min) x (new_max - new_min) + new_min
Z-score: v' = (v - mean) / standard deviation
Decimal scaling: v' = v / 10^j, where j is the smallest integer such that max(|v'|) < 1

Example (min-max to [0, 1]): values 200, 300, 400, 600, 1000: min = 200, max = 1000. The value 400 becomes (400 - 200)/(1000 - 200) = 0.25.

Binning example

Sorted data: 4, 8, 15, 21, 21, 24, 25, 28, 34. Equal-depth bins of size 3: {4, 8, 15}, {21, 21, 24}, {25, 28, 34}. Smoothing by bin means: {9, 9, 9}, {22, 22, 22}, {29, 29, 29}.

Why it matters

Preprocessing improves accuracy, efficiency and reliability of the mining results (garbage in, garbage out) and often takes most of the project effort.

Q9c. Write a short note on Efficient Frequent Itemset Mining. _(Q.9: short notes on any two of the following)_20267m

Module 2: Data Mining Fundamentals and Association Rule Mining

Write a short note on Efficient Frequent Itemset Mining.

(Q.9: short notes on any two of the following)

View this question on its own page →

Search Google Ask AI

Worked Solution

Efficient Frequent Itemset Mining

Frequent itemset mining finds sets of items that occur together in at least a minimum number of transactions (min_sup). It is the first and costliest step of association rule mining. With n items there are up to 2^n - 1 itemsets, so efficient methods are essential.

1. Apriori (level-wise, generate-and-test)

Uses the Apriori property: every subset of a frequent itemset is frequent.
Finds L1, then generates candidates Ck from Lk-1 (join), prunes using the property, and counts support by scanning the database; repeats until no more frequent itemsets.
Drawbacks: many candidates and many database scans.

2. Improvements to Apriori

Technique	Idea
Hash-based itemset counting	Hash k-itemsets into buckets; a bucket with a count below min_sup cannot contain frequent itemsets, so those candidates are dropped
Transaction reduction	A transaction with no frequent k-itemset is not needed in later scans
Partitioning	Split the database; a frequent itemset must be frequent in at least one partition; only 2 scans needed
Sampling	Mine a random sample with a lower threshold; verify on the full data
Dynamic itemset counting	Add new candidates while scanning, at checkpoints, to reduce scans

3. FP-Growth (pattern growth, no candidate generation)

Scan the database once to find frequent items and sort them in descending support.
Scan again to build a compact FP-tree (a prefix tree with a header table of node links).
For each frequent item, build its conditional pattern base and conditional FP-tree, and mine it recursively (divide-and-conquer).

Advantages: only two database scans, no candidate generation, and the tree compresses the data. It is much faster than Apriori on large or dense data.

4. Eclat (vertical data format)

Stores each item with its transaction-ID list (TID-list). The support of an itemset is found by intersecting TID-lists, so no repeated database scans are needed.

5. Closed and maximal itemsets

Mining only closed or maximal frequent itemsets gives a compact summary of all frequent itemsets.

Comparison

Method	Candidates	Scans	Best for
Apriori	Yes (many)	One per level	Small data, simple implementation
FP-Growth	No	2	Large and dense data
Eclat	Yes (via TID-lists)	Few	Sparse data that fits in memory

Conclusion

Efficiency comes from pruning with the Apriori property, reducing database scans, or avoiding candidate generation altogether, as FP-Growth does.

Q9d. Write a short note on Categorization of Major Clustering Methods. _(Q.9: short notes on any two of the following)_20267m

Module 4: Cluster Analysis

Write a short note on Categorization of Major Clustering Methods.

(Q.9: short notes on any two of the following)

View this question on its own page →

Search Google Ask AI

Worked Solution

Categorization of Major Clustering Methods

Clustering groups data objects so that objects in the same cluster are similar to each other and different from objects in other clusters. It is unsupervised learning. The major methods are grouped into five categories.

1. Partitioning methods

Divide n objects into k partitions (clusters), each object belonging to one cluster; the user gives k.

K-means: each cluster is represented by its mean (centroid); objects are repeatedly assigned to the nearest centroid.
K-medoids (PAM, CLARA, CLARANS): each cluster is represented by an actual object (medoid); less sensitive to outliers.
Good for spherical clusters of similar size and for large data; needs k.

2. Hierarchical methods

Build a tree of clusters (dendrogram).

Agglomerative (bottom-up): start with each object as a cluster and merge the closest.
Divisive (top-down): start with one cluster and split.
Linkage: single, complete, average. Variants: BIRCH, CURE, Chameleon.
No k needed in advance, but a merge or split cannot be undone and it is costly for big data.

3. Density-based methods

A cluster is a dense region of objects separated by regions of low density.

DBSCAN (parameters eps and MinPts), OPTICS, DENCLUE.
Finds arbitrary-shaped clusters and handles noise; k not needed; struggles with varying densities.

4. Grid-based methods

Quantize the space into a finite number of grid cells and cluster the cells.

STING, CLIQUE, WaveCluster.
Very fast: the time depends on the number of cells, not on the number of objects.

5. Model-based methods

Assume a mathematical model for each cluster and fit the data to it.

EM (Expectation-Maximization) with Gaussian mixtures, COBWEB (conceptual clustering), SOM (self-organizing maps).
Gives soft (probabilistic) membership; depends on the assumed model.

Summary table

Category	Basic idea	Examples	Strength	Weakness
Partitioning	Split into k groups, iteratively optimize	K-means, K-medoids	Simple, scalable	Needs k, spherical shapes, outliers
Hierarchical	Tree of merges/splits	AGNES, DIANA, BIRCH	Dendrogram, no k	Not scalable, no undo
Density-based	Dense regions	DBSCAN, OPTICS	Arbitrary shapes, noise	Parameter sensitivity
Grid-based	Grid cells	STING, CLIQUE	Fast	Less accurate boundaries
Model-based	Statistical model fit	EM, COBWEB, SOM	Soft clusters	Model assumptions

Some methods combine ideas from several categories (for example high-dimensional and constraint-based clustering).

Conclusion

The right method depends on data size, cluster shape, presence of noise and whether k is known.

Previous year2025

Official AppAndroid App

PYQDeck for Android

Solve BEU previous-year exam papers anywhere, anytime. Fast search, offline paper viewing, and step-by-step solutions designed specifically for your phone.

Get on Google Play

Data Mining

Solution

Explanation

Why other options are incorrect:

Solution

Explanation

Why other options are incorrect:

Key Points to Remember:

Solution

Explanation

Why other options are incorrect:

Solution

Explanation

Why other options are incorrect:

Solution

Explanation

Why other options are incorrect:

Key Points to Remember:

Solution

Explanation

Why other options are incorrect:

Solution

Explanation

Why other options are incorrect:

Solution

Explanation

Why other options are incorrect:

Solution

Explanation

Why other options are incorrect:

Solution

Explanation

Why other options are incorrect:

Pocketly on Android

Components of a Data Warehouse and Their Functions

Architecture

Major components

Functions in the warehouse environment

Conclusion

Data Cube: Sales by Region, Month and Product Category

(i) Total number of possible multidimensional cells

(ii) Cells required for one region across all months and products

(iii) OLAP operations on this cube

Extra: number of cuboids

Conclusion

Preprocessing Pipeline for a Dataset with Numerical, Categorical and Missing Attributes

Pipeline

Why this order

Example

Conclusion

Functionalities of Data Mining and the Role of Preprocessing

Major functionalities

How preprocessing improves quality and reliability

Effects on the discovered knowledge

Conclusion

Apriori Principle, Candidate Generation and Pruning, and Comparison with FP-Growth

Apriori principle

How Apriori works (level-wise search)

How candidate generation and pruning reduce the search space

Apriori vs FP-Growth

Conclusion

Types of Association Rules and When Each Is Useful

Types of association rules

Related types (for completeness)

Situations where each helps

Evaluating the rules

Conclusion

Experimental Framework for Evaluating a Classification Model

Step 1: Prepare the data

Step 2: Training and testing datasets

Step 3: Cross-validation

Step 4: Confusion matrix

Step 5: Evaluation measures

Step 6: ROC-based evaluation

Step 7: Compare and report

How ensemble methods improve predictive performance

Conclusion

Clustering Framework for Customer Segmentation

Framework overview

1. Data and business goal