Info > Data Mining > 3-1. DM and DW > The Architecture of Data

¢¹¢º The Architecture of Data

¢º Data is the heart of both DW and DM

¢º Description way

¢¹ simply differentiation between internal data and external data
¢¹ a more useful way of looking at data is in terms of abstraction

¡Ø operational data - who, what, where and when
¡Ø summary data - summaries by who, what, where, when
¡Ø database schema - physical layout of the data, tables, fields.
¡Ø metadata - logical model and mapping to physical layout and sources.
¡Ø business rule - what's been learned from the data.

¢º Operation Data

¡Ø the most basic form of data

¢¹ Example : every product purchased by a customer, every bank transaction, every credit card purchases, every flight segment, every package, every telephone call.
¢¹ Changes to operational data are quite common
¢¹ The fact that operational data changes has to be part of any robust data warehousing approach.
¢¹ the amount of data : enormous

¡Ø a typical fast food restaurant sells hundreds of thousands of meal over the course of a year.
¡Ø a chain of food markets

¢º Summary Data

¡Ø the first level of abstraction
¢¹ derived from operational data and the most common way that users interact with data
¢¹ summary data is the data in reports that provide static views of the business.
¢¹ today's summary data may not be important in the future.

¢º Database Schema

¡Ø data about the data itself
¢¹ This most basic level is the physical layout of the operational data and the summary data ¡æ tell us what we have and what we can know.
¢¹ Do we know the telephone numbers of customer ?
¢¹ If there is some telephone number field somewhere in the data. Can we determine sales trends for products in different regions?
¢¹ If there are tables with historical sales data for products in different regions or tables with operational data that can be used to create the appropriate summaries.

¢º Metadata

¡Ø a logical data model that defines the data in terms of entities (the product of hierarchy), attributes, and relationships (business and its various partners) meaningful on the business level.
¢¹ The importence of metadata

¡Ø provides the key link between the business users and the data
¡Ø a good metadata system gives users the ability to browse through the metadata on their desktop.
¡Ø accurate and accessible

¢º The business rules

¡Ø not only describe the structure of the data(metadata) but they also describe why relationship exist and how they are applied.
¢¹ a close relationship to DM
¢¹ Example

¡Ø market basket analysis
¡Ø decision tree
¡Ø produce explicit rules ¡æ business ¡æ DM technique

¢¹ Even more generally, the process of using DM should be captured as business rules as well as specific results

Info > Data Mining > 3-1. DM and DW > The Architecture of Data