Q.

1 Attempt the following (any 5) (25-10M)

i. Star Schema: It is a data modeling technique in which a central "fact"
table is connected to multiple "dimension" tables, forming a star-like
structure. Snowflake Schema: It is an extension of the star schema
where dimension tables are further normalized into multiple levels,
creating a snowflake-like structure.

ii. Table Inheritance: It is a concept in database design where one table

inherits attributes and relationships from another table, forming a
hierarchy. For example, a "Person" table may inherit attributes from a
"Customer" table, which in turn inherits from a "User" table, creating a
hierarchy of tables with increasing specialization.

iii. MOLAP (Multidimensional Online Analytical Processing): It is a type

of database technology that stores data in a multidimensional cube
format. It allows for efficient analysis of large volumes of data and
enables complex calculations, aggregations, and slicing/dicing
operations for analytical purposes.

iv. Dendrograms require: 1) Distance or similarity measures between

objects, 2) A method for merging or splitting clusters, 3) A stopping
criterion to determine when to stop the clustering process, and 4) A
visualization method to represent the hierarchical structure of clusters.

v. Advantages of parallel databases include: 1) Increased performance

and scalability through parallel processing, 2) High availability and fault
tolerance, 3) Efficient utilization of hardware resources, and 4) Support
for complex analytical queries involving large datasets.

vi. Uses of data warehouses: 1) Business intelligence and decision-

making, 2) Trend analysis and forecasting, 3) Customer relationship
management (CRM) and market segmentation, 4) Performance
measurement and monitoring, and 5) Data mining and advanced
analytics.

vii. Data mining refers to the process of discovering patterns,

relationships, and insights from large datasets. It involves extracting
valuable information, knowledge, or patterns that were previously
unknown, and using them to make informed decisions or predictions.
Data mining techniques include classification, clustering, regression,
and association rule mining.

Q.2 Answer the following (any 3): (4*3=12M)

i. Parallel query evaluation refers to the execution of database queries
by dividing the workload among multiple processors or nodes. It aims to
improve query performance by processing data in parallel, reducing
response times. Parallelism can be achieved through techniques such as
parallel query execution plans, parallel data processing, and parallel
data distribution.
ii. Data warehousing is needed to address the limitations of traditional
transactional databases when it comes to analyzing large volumes of
data. It provides a centralized repository for storing, integrating, and
managing data from various sources. Data warehouses enable efficient
data retrieval, complex analysis, and reporting, supporting decision-
making processes and business intelligence initiatives.

iii. Online Analytical Processing (OLAP) is a technology used for data

analysis and reporting. It allows users to perform multidimensional
analysis on large volumes of data, providing a flexible and intuitive way
to explore information. OLAP systems provide features like drill-down,
roll-up, slicing, and dicing to analyze data from different perspectives,
enabling users to gain insights and make informed decisions.

iv. Different distance formulas are used to calculate the similarity or

dissimilarity between objects or data points in various data mining and
clustering algorithms. Some commonly used distance formulas include
Euclidean distance, Manhattan distance, Chebyshev distance,
Minkowski distance, and Hamming distance. Each formula has its own
characteristics and is suitable for different types of data and analysis
scenarios.

v. Implementing Object-Relational Database Management Systems

(ORDBMS) introduces new challenges, such as: 1) Data modeling
complexities due to the integration of object-oriented features, 2)
Performance optimization for object-oriented queries and complex
relationships, 3) Incorporating object-oriented programming paradigms
into the database system, 4) Managing the evolution and maintenance
of object schemas, and 5) Ensuring compatibility with existing relational
database systems and standards. Overcoming these challenges requires
careful planning, design, and implementation strategies.

Q.3 Answer the following (any 2) (2*5=10M)

i. Parallel databases and distributed databases are both designed to
handle large amounts of data, but they differ in their approach:
- Parallel Databases: In a parallel database, a single database is divided
across multiple processors or nodes, and each node processes a portion
of the data simultaneously. The nodes communicate and coordinate
their actions to execute queries in parallel, improving performance.
- Distributed Databases: In a distributed database, data is physically
distributed across multiple nodes or sites, which can be geographically
dispersed. Each node maintains its own subset of the data and operates
autonomously. Queries may be executed locally or distributed across
multiple nodes for processing.

ii. The drill down feature in OLAP allows users to navigate from
summarized data to more detailed levels of information. For example,
consider an OLAP cube representing sales data. Initially, the user may
view the total sales for a specific region. By drilling down, the user can
access more detailed information, such as sales by city, and further drill
down to sales by store or individual products.

iii. Complex data types in SQL refer to the ability to store structured or
semi-structured data within a relational database. Examples include:
- Arrays: A collection of values of the same type.
- Structs: A composite data type that groups multiple related fields
together.
- JSON: Support for storing and querying JSON (JavaScript Object
Notation) data.
- XML: Support for storing and manipulating XML (eXtensible Markup
Language) data.
- Spatial data types: Allows for storing and querying geographic or
geometric data.
- User-defined types: The ability to define custom data types based on
specific requirements.

iv. To divide the given dataset into two clusters using the k-means
algorithm, an initial step is to randomly assign two cluster centers, let's
say C1 and C2. Then, each data point is assigned to the cluster whose
center it is closest to. The mean of the data points in each cluster is
computed, and the cluster centers are updated accordingly. This process
is iteratively repeated until the cluster assignments stabilize. The
resulting clusters for the given dataset may vary depending on the
initialization and the distance metric used.

Q.4 Answer the following (any 2) (2*6-12M)

i. Architecture of a Data Warehouse:

A data warehouse typically follows a three-tier architecture, consisting

of the following components:
1. Source Systems: These are the systems that generate and store
operational data, such as transactional databases, CRM systems, or
external data sources. Data is extracted from these source systems and
transformed before being loaded into the data warehouse.

2. Data Warehouse: The data warehouse serves as the central

repository for consolidated and integrated data. It consists of multiple
components:

- Staging Area: It is the initial landing zone where data is extracted

from source systems. Data is stored temporarily in its raw form before
undergoing transformation.

- Data Integration Layer: This layer performs the extraction,

transformation, and loading (ETL) process. Data is cleansed,
standardized, and transformed to match the data warehouse schema.

- Data Storage: This component stores the transformed and structured

data. It typically uses a relational database management system
(RDBMS) or a columnar database to support efficient querying and
analysis.

- Metadata Repository: It stores the metadata, which provides

information about the data in the data warehouse, including its
structure, source, and transformation rules.
3. Business Intelligence (BI) Layer: This layer provides tools and
interfaces for data analysis, reporting, and visualization. Users can
access and interact with the data warehouse through various BI tools,
such as dashboards, ad hoc query tools, or OLAP tools.

Here is a simplified diagram representing the architecture of a data

warehouse:

```
+---------------------+
| Business Intelligence |
| (BI) Layer |
+---------------------+
|
v
+---------------------+
| Data Storage |
| and Processing |
+---------------------+
|
v
+--------------+--------------+---------------------+
 | Staging | Data | Metadata |
| Area | Integration | Repository |
+--------------+--------------+---------------------+
|
v
+---------------------+
| Source Systems |
+---------------------+
```

ii. Agglomerative Clustering:

Agglomerative clustering is a hierarchical clustering technique that

starts with each data point as an individual cluster and then iteratively
merges the closest clusters until a stopping criterion is met. It follows
these steps:

1. Initialization: Each data point is initially considered as a separate

cluster.

2. Distance Calculation: The distance between each pair of clusters is

computed based on a chosen distance metric (e.g., Euclidean distance).
3. Merge: The two closest clusters are merged into a single cluster. The
distance between clusters is updated using linkage methods like single
linkage, complete linkage, or average linkage.

4. Repeat: Steps 2 and 3 are repeated until a stopping criterion is met.

This could be a predetermined number of clusters or a specific level of
similarity.

5. Create Dendrogram: A dendrogram is a visual representation of the

clustering process, showing the hierarchy of merged clusters.

Agglomerative clustering is a bottom-up approach, where small clusters

are progressively merged into larger ones. The result is a hierarchical
tree-like structure that can be cut at different levels to obtain clusters of
desired sizes.

iii. Bayesian Classification Technique:

Bayesian classification is a statistical technique used for supervised

classification tasks. It utilizes Bayes' theorem to estimate the probability
of a data point belonging to a particular class based on its observed
features.

The technique involves the following steps:

1. Training: During the training phase, the algorithm learns the
underlying probability distribution of each class and the relationships
between features. It calculates the prior probabilities of each class and
the likelihood of each feature given the class.

2. Prediction: In the prediction phase, the algorithm applies

Bayes' theorem to estimate the posterior probability of each class given

the observed features. It selects the class with the highest posterior
probability as the predicted class for the data point.

Bayesian classification assumes that features are conditionally

independent given the class. Naive Bayes classifier is a popular variant
that makes this assumption, simplifying the calculations. It is widely
used for text classification, spam filtering, and other applications where
probabilistic reasoning is required.

iv. Structured Types and Inheritance in SQL:

Structured types and inheritance are features in SQL that allow for the
definition of custom data types and relationships between them:

1. Structured Types: SQL supports the creation of user-defined

structured types, also known as composite types or object types. These
types can consist of multiple attributes, each with its own data type. For
example, a "Person" structured type may have attributes like "name,"
"age," and "address." Structured types provide a way to encapsulate
related attributes into a single unit.

2. Inheritance: SQL also supports the concept of inheritance between

structured types. Inheritance allows one structured type to inherit
attributes and behavior from another type, forming a hierarchy. For
example, a "Student" structured type can inherit attributes from the
"Person" type, adding additional attributes specific to students.
Inheritance allows for code reuse, data organization, and polymorphism
within the database.

These features provide flexibility and extensibility in data modeling,

allowing developers to define custom data structures and relationships
based on specific requirements.

Q.5 Answer the following (any 2) (2*8=16)

i. Certainly! Here's an example of a distance matrix for a

dataset with five data points (A, B, C, D, E) using the divisive
clustering technique:

A B C D E
A 0 5 3 4 6
B 5 0 2 3 7
C 3 2 0 6 4
D 4 3 6 0 5
E 6 7 4 5 0

In this example, the distance matrix represents the dissimilarities or

distances between each pair of data points. The values in the matrix are
the distances between the respective data points, calculated using a
chosen distance metric (e.g., Euclidean distance, Manhattan distance,
etc.).

For instance, the distance between data point A and data point B is 5,
between A and C is 3, between A and D is 4, and so on.

This distance matrix can be used as input for the divisive clustering
algorithm to perform the clustering process, as explained in the
previous response.

ii. Certainly! Here's an example dataset consisting of four data

points (A, B, C, D):

A: (2, 3)
B: (5, 1)
C: (4, 6)
D: (7, 2)
To create a distance matrix using the Euclidean formula, we calculate
the Euclidean distance between each pair of data points.

The Euclidean distance formula between two points (x1, y1) and (x2, y2)
is:

Distance = sqrt((x2 - x1)^2 + (y2 - y1)^2)

Using this formula, we can calculate the distance between each pair of
points:

Distance between A and B:

sqrt((5 - 2)^2 + (1 - 3)^2) = sqrt(9 + 4) = sqrt(13) ≈ 3.61

Distance between A and C:

sqrt((4 - 2)^2 + (6 - 3)^2) = sqrt(4 + 9) = sqrt(13) ≈ 3.61

Distance between A and D:

sqrt((7 - 2)^2 + (2 - 3)^2) = sqrt(25 + 1) = sqrt(26) ≈ 5.10

Distance between B and C:

sqrt((4 - 5)^2 + (6 - 1)^2) = sqrt(1 + 25) = sqrt(26) ≈ 5.10
Distance between B and D:
sqrt((7 - 5)^2 + (2 - 1)^2) = sqrt(4 + 1) = sqrt(5) ≈ 2.24

Distance between C and D:

sqrt((7 - 4)^2 + (2 - 6)^2) = sqrt(9 + 16) = sqrt(25) = 5.00

Now we can arrange these distances in a distance matrix:

This distance matrix represents the Euclidean distances between each

pair of data points. It can be used as input for various data mining or
clustering algorithms that rely on distance-based computations.
iii. The k-medoid algorithm is a clustering algorithm that aims to
partition a dataset into k clusters, where each cluster is represented by
a medoid, which is the most centrally located data point within the
cluster. It is a variation of the k-means algorithm that uses actual data
points as cluster centers instead of the mean.

The k-medoid algorithm follows these steps:

1. Initialization: Randomly select k data points from the dataset as the

initial medoids.

2. Assignment: Assign each data point to the nearest medoid based on a

chosen distance metric, such as Euclidean distance or Manhattan
distance.

3. Medoid Update: For each cluster, calculate the total dissimilarity or

distance between each data point in the cluster and all other data
points. Select the data point with the lowest total dissimilarity as the
new medoid for that cluster.

4. Iteration: Repeat steps 2 and 3 until the medoids no longer change or

a stopping criterion is met. The stopping criterion could be a
predetermined number of iterations or a threshold for the change in
medoids.
The k-medoid algorithm is less sensitive to outliers compared to k-
means since medoids are actual data points from the dataset. However,
it can be computationally expensive, especially for large datasets, as it
requires calculating the dissimilarity between all pairs of data points.

Example:
Let's consider a dataset of customers with their annual income and
spending score. We want to cluster them into two groups using the k-
medoid algorithm.

Dataset:
Customer 1: (Income = $50,000, Spending Score = 60)
Customer 2: (Income = $30,000, Spending Score = 40)
Customer 3: (Income = $70,000, Spending Score = 90)
Customer 4: (Income = $80,000, Spending Score = 70)
Customer 5: (Income = $20,000, Spending Score = 20)

1. Initialization: Randomly select two data points as initial medoids:

Customer 1 and Customer 5.

2. Assignment: Calculate the distance between each data point and the
medoids. Assign each data point to the nearest medoid.

Cluster 1: {Customer 1, Customer 2, Customer 4}

 Cluster 2: {Customer 3, Customer 5}

3. Medoid Update: Calculate the total dissimilarity of each data point

within its cluster and select the data point with the lowest total
dissimilarity as the new medoid.

Cluster 1: Customer 2 (new medoid)

Cluster 2: Customer 5 (new medoid)

4. Iteration: Repeat steps 2 and 3 until the medoids no longer change.

After further iterations, the algorithm converges, and the final

clustering result is:

Cluster 1: {Customer 2, Customer 4}

Cluster 2: {Customer 1, Customer 3, Customer 5}

The medoids Customer 2 and Customer 5 represent their respective

clusters, and the other customers are assigned to the nearest medoid
based on distance.

iv. KDD (Knowledge Discovery in Databases) is a process used to extract

useful knowledge or insights from large volumes of data. It involves
several iterative steps:
1. Problem Definition: Clearly define the goal and objectives of the
knowledge discovery process, along with the specific data mining tasks
to be performed.

2. Data Preparation: Gather and preprocess the data required for

analysis. This involves data selection, cleaning, integration, and
transformation to ensure data quality and compatibility.

3. Data Mining: Apply various data mining techniques, such as

clustering, classification, regression, association rules, or anomaly
detection, to extract patterns, relationships

, and insights from the prepared data.

4. Pattern Evaluation: Evaluate and interpret the patterns and models

generated by the data mining algorithms. Assess their validity,
usefulness, and reliability in addressing the defined problem.

5. Knowledge Presentation: Present the discovered knowledge and

insights in a meaningful and understandable format to stakeholders.
This may involve visualization, reports, dashboards, or interactive tools
for decision-making.
6. Knowledge Utilization: Apply the discovered knowledge and insights
to make informed decisions, develop strategies, or improve processes in
the relevant domain.

The KDD process is iterative, and feedback from stakeholders and

domain experts is crucial at each step. It involves a combination of
domain knowledge, statistical analysis, machine learning, and data
visualization techniques to uncover hidden patterns and gain valuable
insights from data.