- Evolution of decision supportDifferent categories of DS analysisSimple queries….. More advanced analysis techniquesMost widespread analysis technique
Go from simple queries to advanced analysis
- Simple consultation: fresh data, strategic decisions "canned" or predefined consultations Consultations without preparation (ad-hoc)
MULTIDIMENSIONAL ANALYSIS:
The power to divide
Standard query multidimensional analysis
More powerful analysis technique
- Delve deeper
Different perspectives of data by using “dimensions”
Information by a certain dimension
Drill down
Similar tools (to standard query)
Same data: different ways
Modeling and Segmentation:
Analysis for knowledge workers
Bulky and detailed data
Knowledge workers
Model. Collection of patterns for a given characteristic
Modeling algorithms Decision support
Segmentation is also not a NEW tool
Example: CLIENT SEGMENTATION
Specialized software
Knowledge Discovery
- Marketing looking for the best clientsKD. Powerful algorithms looking for patterns in large DB Patterns not specified beforehand (as in modeling) DW tells company where important patterns and relationships are “Unknown Answers” to find new innovative ways
Some observations
- Evolution of the DW infrastructureEvolution of the DW infrastructureData in DW evolve and matureDW as a basic need
Increased frequency of use
Time or cost savings
Process consolidation
Standard Queries
“List the # of loans made to customer X and the payment dates when it has taken more than a week”
The consultation seeks to confirm an alleged part of a very strong hypothesis
"List all customers for which the revenue from use at peak time has decreased by 20% or more" Result commercial actions
"Show all cell phone customers with incomplete incoming calls by more than 20% in each week"
Useful results:
- High-use customers needing additional equipment Potential market competitors Candidates for new product offerings Fraud suspicions Potentially dissatisfied customers In short, retain and add customers
"Show all cell phone customers with incomplete incoming calls by more than 20% per week" Drill-down
“Of those clients, who use roaming outside the network?
Results:
"Show the quarterly income corresponding to large commercial clients in the northern, NW and SO regions in 1977 and 1978"
“Show the same data by district within the Northwest region”
“Show the same data for District B in the Northwest region”
OLAP tools are unique to facilitate the user to request the same data in different ways
SUPPLIERS
- Oracle OLAPSAP - OLAPMicrosoftInesoftDundas Chart for.NET OLAP ServicesMicroStrategy OLAP Services
Modeling
- Future customer behavior Your long-term viability
Segmentation
- Sort and reclassify customers by Demography Purchase patterns Proclivity to purchase
Modeling Examples
- Customer lifetime value Customer wear Predictive modeling
Segment Examples
- Customers who respond to offers Customers who respond to discounts Customers who respond to new products Customers who respond to promotions
Segmentation results:
- Which group of customers to target a new service? Who is most likely to be interested in this service? Which customers are most likely to commit fraud? Which customers are most likely to respond to discounts? Zero hypothesis Find in the pattern data Hidden: Customer Behavior Sales by Product Cancellations Future Purchases
Examples:
- What happens when 2 drugs are taken together Unknown product affinity patterns Other trigger products
More examples:
Of customers who buy fried foods, 66% also buy sweets
Of the customers who buy fried foods and sweets, 75% also buy red wine
ACTIONS
- Send coupons to frequent shoppers Eliminate discounts on red wine Relocate products in-store Examples of data that come to light thanks to KD: Certain trigger products affect other purchases Discover the “next likely purchase” Discover patterns in the fall of purchases Customer aspects that affect the product life cycle Data mining comprises segmentation, modeling and knowledge discovery It is synonymous with statistical analysis It is supported by 3 technologies Massive data collection Powerful computers with multiprocessors Data Mining algorithms
It is developed under the latest generation languages based on artificial intelligence and uses mathematical models such as:
- Artificial Neural Networks Decision Trees Induction Rules Genetic Algorithms
Predicts future trends and behaviors to enable businesses to make proactive decisions
These tools can answer time-consuming business questions
They scan databases for hidden patterns, finding predictable information that an expert cannot find because it is outside of their expectations.
Data Warehouse in reality
The data resides in the least amount of different platforms, preferably in a DW or Data mart
Managers usually access the DW directly when they require information
Executives support decision support
DW can improve productivity and increase profitability of the organization
References
E-data. Transforming data into information with Data Warehousing
Chapter 2. Supporting the Bottom-Up Decision Jill Dyché
Download the original file