HND - BI - W9 - Multidimensional Data Models and Analysis

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

Unit 14 business intelligence

6 Multidimensional Data Models and Analysis

1
Why Data Model?

▪ To make better decisions faster, knowledge workers required


useful information and those information flexible enough to
analysis.

▪ The problem many organization faced was how to convert these


raw data gathered in to information that ready for analysis.

▪ Responsibility of Business Intelligent system (BI) to convert these


raw data to flexible easy to analyze information by properly
structuring and presenting them using various data models.

2
Raw Data…. Where They Come From ?

▪ Raw data come to the organization via operational data


sources. These may include
▪ OLTP systems.
▪ File systems and individual files
▪ Legacy systems.
▪ External data sources

3
Raw Data… Why We Can’t Analyze

▪ OLTP data structures are not efficient for answer analysis


queries
▪ The raw data available OLTP is not very useful for analysis.
▪ Limited/pre made reporting and visualization options.
▪ Not easy to handle the calculation, sorting, summarizing .
▪ Analysis queries reduce OLTP system performance.
▪ System contain non aggregated, system specific data (Data
Silos)
▪ OLTP does not support Multi dimensional data analysis

4
OLAP

E.F Codd inventor of relational model coined the term


OLAP. He gives three must have capabilities OLAP
Systems.
▪ Must support multidimensional analysis.
▪ Fast information retrieval time
▪ Must have robust calculation engine to handle calculations
efficiently across the all dimension

5
OLAP Key Words

▪ Dimension
All the members within the domain that can be group/aggregate
together or compared together.
▪ Bike Brand Honda, Yamaha, Suzuki, Kawasaki measurement
can compared together
▪ January ,February ,March can compare each other and
aggregate for 1st Quarter,Four Quarters can aggrigate to Year
and so on
▪ Measurement:
Any quantitative expression or value that analyzing in OLAP
system
▪ Sales value, Sales quantity , Number of registration

6
Data Cube or "Multi-Dimensional" Data Stores

▪ A long time a go sales data present as a sheet of paper.


Branch

Product

▪ Over the time each and every month add new sheet and
now it is a stack by creating cube
Branch
Product
Month

7
Multi-Dimensional Analysis

▪ This concept of data cube is


▪ Natural.
▪ Easy to understand.
▪ Effective. When analyzing data
▪ Even before BI and OLAP system proposed analyst and
managers interpret data in multi-dimensionally
▪ What was the total value of sales by branch by brand by each quarter.
▪ Each and every “By” qualify as Dimension

8
Example-Nippon Bike Shop
(Not a real data set)

9
Example-Nippon Bike Shop Aggregated Data
(Not a real data set and definitely not accurate)

10
Example-Nippon Bike Shop

Analyst Can easily convert


pervious report in to this
model
▪ Axis/Dimension.
▪ Branch
▪ Brand
▪ Time(Quarter ,Month, Year)
▪ Values/Measurement.
▪ Sales QTY
▪ Sales value ($)

11
Data Cubes

▪ Even though called this


structure data cube, the data
structure can add n… number
of dimensions. The correct
name is multidimensional
database.
▪ Could be store in computer
hard drive or RAM memory

12
Multi Dimensional Benefits

▪ Multi dimensional, Data structure can handle n.. number of


dimensions. Because of that analysis can conduct over
large number of dimensions.
▪ Simple structure for user to understand.
▪ Can be extract and analyze data effectively via pre-
structured and aggregated data
▪ Drill down achieved either within the cube or through other
cubes of lesser grain, or by accessing the Data Warehouse

13
Multi Dimensional Drawback

• Pre-design and storage of data necessary for the


Cubes
• Because of creating separate physical or logical data
base it is consume huge computer resources
• Rather inflexible as regards non pre-defined
requirements
• Data extraction gives operational overhead.
• Weaknesses in scarcity handling, can lead to increased
cube sizes

14
Multi Dimensional Operation
(Slicing Data Cube)
▪ Slicing Data Cube men
selecting values from one
dimension.
▪ All sales from Kurunegala branch
by brand and by quarter.
▪ All sales of Yamaha motor bikes
by branch and by quarter.
▪ Example query
Select Branch, Quarter, count(*)
From Sales
Where Brand= ‘Yamaha ’
Group by Branch, Quarter

15
Multi Dimensional Operation
(Dicing Data Cube)
▪ Select Values from two or
more dimension.
▪ Honda Bike sales In
Kurunegala.
▪ 1st Quarter sales in Colombo
Branch
▪ Example query
Select Quarter, Count( *)
From Sales
Where Brand= ‘Yamaha’ and Branch =‘Colombo’
Group by Quarter

16
Multi Dimensional Operation
(Drill down data cube)
▪ When knowledge worker required more detail view of data
they can drill down data cube .
▪ For example Nippon Bike shop data cube created for
quarterly sales values but if someone required to access
monthly or daily sales that is also possible if data is
available in data warehouse
▪ Example query
Select Year ,Month,Sum (sales_values)
From sales
Group by Year, Month

17
Multi Dimensional Operation
(Roll up cube)
▪ When knowledge worker required more abstract view of
data they can Roll up or drill up data cube.
▪ After successfully running years If top management in
Nippon Bike shop management required yearly sales
figures. Then they should roll up sales values
▪ Example query
Select Year ,Sum (sales_values)
From sales
Group by Year

18
Multi dimensional Operation
(Pivoting)
▪ Pivoting is the operation that re-organizing dimension to
change the view of data cube.
▪ Pivoting helps to make an effective data visualization by
re arranging X,Y,Z axis

19
Multi dimensional Operation
(Pivoting)
Table Pivoted by quarterly sales by Branch By Brand

Honda Yamaha Suzuki Kawasaki


Q1 Kurunegala 150 200 250 600
Colombo 200 100 50 250
Mathara 150 300 400 450
Kandy 450 50 200 100
Q2 Kurunegala 50 450 100 200
Colombo 450 300 400 150
Mathara 100 200 250 50
Kandy 250 600 150 200
Q3 Kurunegala 200 100 50 250
Colombo 150 200 250 600
Mathara 450 50 200 100
Kandy 150 300 400 450
Q4 Kurunegala 450 300 400 150
Colombo 50 450 100 200
Mathara 250 600 150 200
Kandy 100 200 250 50

20
Multi dimensional Operation
(Pivoting)
Table Pivoted by Branch sales by quarter by brand
Honda Yamaha Suzuki Kawasaki
Kurunegala Q1 150 200 250 600
Q2 50 450 100 200
Q3 200 100 50 250
Q4 450 300 400 150
Colombo Q1 200 100 50 250
Q2 450 300 400 150
Q3 150 200 250 600
Q4 50 450 100 200
Mathara Q1 150 300 400 450
Q2 100 200 250 50
Q3 450 50 200 100
Q4 250 600 150 200
Kandy Q1 450 50 200 100
Q2 250 600 150 200
Q3 150 300 400 450
Q4 100 200 250 50

21
Bibliography

▪ Lamont, M. (2014). Business Intelligence: Multidimensional


Analysis. [online] YouTube. Available at:
https://www.youtube.com/watch?v=IhFkNmVmwn4&t=1285
s /> [Last Accessed 2 May 2020].
▪ Boyer, J. (2010) Business Intelligence Strategy. MC Press
(US).
▪ Marr, B. (2015) Big Data: Using Smart Big Data, Analytics
and Metrics to Make Better Decisions and Improve
Performance. 1st Ed. John Wiley & Sons, Ltd.

22
Next Week

▪ Implementing BI Strategy and BI Systems


▪ What is Business Intelligence Strategy?
▪ Why Organizations Need BI System
▪ Creating Effective BI strategy

23

You might also like