Chapter 1: Introduction to Data Mining
1. What is data mining? Discuss its importance and how it differs from
traditional data analysis techniques.
Answer:
Data mining is the process of discovering hidden patterns, relationships, and useful information from
large datasets using statistical, machine learning, and database techniques.
Importance:
● Helps in decision-making
● Identifies trends and patterns
● Enables prediction of future outcomes
● Improves business operations (marketing, fraud detection, healthcare analytics)
● Extracts meaningful insights from huge datasets
Difference from traditional data analysis:
Traditional Data Analysis Data Mining
Focuses on verifying known hypotheses Focuses on discovering unknown patterns
Manual analysis Automated or semi-automated
Works on small datasets Works on very large datasets
Uses simple statistics Uses AI, ML, and advanced algorithms
2. Explain the major steps involved in the data mining process with an
example.
Answer:
Major steps:
1. Data Cleaning – Removing noise, handling missing values.
2. Data Integration – Combining data from multiple sources.
3. Data Selection – Choosing relevant data for mining.
4. Data Transformation – Normalization, aggregation.
5. Data Mining – Applying algorithms (classification, clustering).
6. Pattern Evaluation – Identifying useful patterns.
7. Knowledge Presentation – Visualizing results.
Example:
A bank wants to find customers who may default loans.
● Clean customer data
● Combine data from credit history and salary records
● Select relevant attributes
● Apply classification algorithms
● Evaluate accuracy
● Present final prediction report
3. What are the various types of data mining tasks? Differentiate between
descriptive and predictive tasks.
Answer:
Types of data mining tasks:
● Classification
● Prediction
● Clustering
● Association rule mining
● Outlier detection
● Summarization
● Sequential pattern mining
Descriptive vs Predictive:
Descriptive Predictive
Describes patterns in existing data Predicts unknown or future data
Includes clustering, association, summarization Includes classification, regression
Answers "What happened?" Answers "What will happen?"
Chapter 2: Data Preprocessing
4. Why is data preprocessing important in data mining? Explain the key steps
involved in data preprocessing.
Answer:
Data preprocessing is important because raw data is often incomplete, inconsistent, noisy, or
unformatted.
Without preprocessing, results may be inaccurate.
Key steps:
1. Data Cleaning – Handling missing values, smoothing noise
2. Data Integration – Combining data from multiple sources
3. Data Transformation – Normalization, aggregation, encoding
4. Data Reduction – Dimensionality reduction, sampling
5. Data Discretization – Converting continuous data into categories
5. Describe the different methods of handling missing data. Provide scenarios
where each method is appropriate.
Answer:
Methods:
1. Ignore the tuple
○ Use when dataset is large and missing values are very few.
2. Fill with global constant
○ Example: Unknown, NA. Useful for categorical data.
3. Fill with mean/median/mode
○ Mean → numerical data
○ Median → skewed data
○ Mode → categorical data
4. Fill with most probable value using prediction
○ Use regression, decision trees, KNN.
○ Suitable for important datasets where accuracy is needed.
6. What are data normalization techniques? Explain any two normalization
methods with suitable examples.
Answer:
Normalization scales data into a smaller, uniform range (commonly 0–1).
It improves model accuracy and prevents bias.
Two methods:
1. Min-Max Normalization
Maps values to range [0,1].
Formula:
v′=v−minmax−minv' = \frac{v - min}{max - min}v′=max−minv−min
Example:
If values = {10, 20, 30}, and you want to normalize 20:
v′=20−1030−10=0.5v' = \frac{20 - 10}{30 - 10} = 0.5v′=30−1020−10=0.5
2. Z-Score Normalization
Based on mean and standard deviation.
Formula:
v′=v−μσv' = \frac{v - \mu}{\sigma}v′=σv−μ
Example:
If value = 50, mean = 40, SD = 5
v′=50−405=2v' = \frac{50 - 40}{5} = 2v′=550−40=2
Chapter 3: Concept Description
7. Define concept description in data mining. How does it help in
understanding large datasets?
Answer:
Concept description is a form of data summarization that provides concise explanations of large
datasets, often using characteristics, rules, and patterns.
How it helps:
● Summarizes large data into understandable information
● Identifies common features of a data class
● Helps compare different groups
● Supports decision-making
Examples include:
● Summaries of customer groups
● Feature descriptions of fraud transactions
8. Differentiate between data characterization and data discrimination with
examples.
Answer:
Data Characterization Data Discrimination
Summarizes general characteristics of a target Compares differences between two or more
class classes
Example: “Customers with high spending usually Example: “High spenders differ from low
belong to age group 25–40.” spenders in income and location.”
Examples:
● Characterization: Average marks of all students in a college.
● Discrimination: Compare performance of boys vs girls.
Chapter 4: Classification and Prediction
9. What is the difference between classification and prediction in data mining?
Provide real-world examples of each.
Answer:
Classification Prediction
Predicts categorical output Predicts continuous output
Example: Spam or Not Spam Example: Estimating house price
Uses methods like Decision Trees, SVM Uses Regression methods
10. Explain the working of a decision tree algorithm. What are the advantages
and disadvantages of using decision trees for classification?
Answer:
A decision tree is a tree-like structure where:
● Root node = best attribute
● Internal nodes = decision tests
● Branches = outcomes
● Leaf nodes = final class label
Working:
1. Select the best attribute using metrics like Information Gain, Gini Index, etc.
2. Split data based on attribute values.
3. Recursively build subtrees.
4. Stop when:
○ All records have same class
○ No attributes left
Advantages: Disadvantages:
● Easy to understand and visualize ● Easily overfits data
● Works on both numerical and
categorical data ● Small changes = different tree
● Requires little data preprocessing
● Fast training ● Not suitable for large number of
attributes without pruning