Transcript Data Warehouse - San Francisco State University

Business Intelligence - 2
BUS 782
Topics
• Data warehousing
• Data Mining
Data Warehouse
• Data warehouse is a central repositories of
integrated data from one or more sources
created for reporting and data analysis.
– sourced from various operational systems in use
in the organization,
– structured in a way to specifically address the
reporting and analytic requirements.
Example:
Transaction Database
CID
Cname
City
OID
ODate
Rating
SalesPerson
Customer
1
Has
M
Order
M
Qty
Has
M
Product
Price
PID
Pname
Analyze Sales Data
Detailed Business Data
• Total sales:
– by product:
• Qty*Price of each detail line
• Sum (Qty*Price)
• Detailed business data: qty*price
• Total quantity sold:
– By product:
• Sum(Qty)
• Detailed business data: Qty
Dimensions for Data Analysis:
Factors relevant to the business data
• Analyze sales by Product
• Analyze sales related to Customer:
– Location: Sales by City
– Customer type: Sales by Rating
• Analyze sales related to Time:
– Quarterly, monthly, yearly Sales
• Analyze sales related to Employee:
– Sales by SalesPerson
Data Warehouse Design
- Star Schema • Dimension tables
– contain descriptions about the subjects of the
business such as customers, employees, locations,
products, time periods, etc.
• Fact table
– contain detailed business data with links to
dimension tables.
Star Schema
Location
Dimension
LocationCode
State
City
Can group by State, City
FactTable
LocationCode
PeriodCode
Rating
PID
Qty
Amount
Product
Dimension
PID
Pname
Category
CustomerRating
Dimension
Rating
Description
Period
Dimension
PeriodCode
Year
Quarter
Define Location Dimension
• Location:
– In the transaction database: City
– In the data warehouse we define Location to be
State, City
• San Francisco -> California, San Francisco
• Los Angeles -> California, Los Angeles
– Define Location Code:
• California, San Francisco -> L1
• California, Los Angeles -> L2
Define Period Dimension
• Period:
– In the transaction database: Odate
– In the data warehouse we define Period to be:
Year, Quarter
• Odate: 11/2/2003 -> 2003, 4
• Odate: 2/28/2003 -> 2003, 1
– Define Period Code:
• 2003, 4 -> 20034
• 2003, 1 -> 20031
The ETL Process
L
T
One,
companywide
warehouse
E
Periodic extraction  data is not completely current in warehouse
The ETL Process
• Capture/Extract
• Transform
– Scrub(data cleansing),derive
– Example:
• City -> LocationCode, State, City
• OrderDate -> PeriodCode, Year, Quarter
• Load and Index
ETL = Extract, transform, and load
Performing Analysis
• Analyze sales:
– by Location
– By Location and Customer Type
– By Location and Period
– By Period and Product
• Pivot Table:
– Drill down, roll up, reaggregation
Data Mining
• Knowledge discovery using a blend of
statistical, artificial intelligence, and
computer graphics techniques
• Goals:
– Explain observed events or conditions
– Explore data for new or unexpected relationships
Typical Data Mining Techniques
•
•
•
•
Statistical regression
Decision tree induction
Clustering – discover subgroups
Affinity – discover things with strong mutual
relationships
• Sequence association – discover cycles of evens and
behaviors
• Rule discovery – search for patterns and correlations
• Text mining (analytics)
Typical Data Mining Applications
• Profiling populations
– High-value customers, credit risks, credit card fraud
•
•
•
•
Analysis of business trends
Target marketing
Campaign effectiveness
Product affinity
– Identifying products that are purchased concurrently
• Up-selling
– Identifying new products and services to sell to a customer
based on critical events
Affinity Analysis:
Market Basket Analysis
• Market Basket Analysis is a modeling technique
based upon the theory that if you buy a certain
group of items, you are more (or less) likely to buy
another group of items.
• The set of items a customer buys is referred to as an
itemset, and market basket analysis seeks to find
relationships between purchases.
• Typically the relationship will be in the form of a rule:
Example:
– IF {beer, no bar meal} THEN {chips}.
Basket Analysis and Cross- Selling
• For instance, customers are very likely to purchase
shampoo and conditioner together, so a retailer
would not put both items on promotion at the same
time. The promotion of one would likely drive sales
of the other.
• A widely used example of cross selling on the
internet with market basket analysis is
Amazon.com's use of suggestions of the type:
– "Customers who bought book A also bought book B", e.g.
Text Mining
• Objective: deriving high-quality information
from text.
– text categorization
– text clustering
– concept/entity extraction
– sentiment analysis, etc.
Social Media Mining
• Salesforce Radian6 Social Marketing Cloud
• http://www.youtube.com/watch?v=EH1dcFh_-I4