Why is creating a data warehouse so difficult?
A data warehouse brings together a wide range of data that may reside in different types of databases, using different norms for entered data and have various states of integrity. Extracting data from these legacy databases may prove difficult as will reconciling the differences between the databases as the data are brought together in the data warehouse.
Do you need to keep meta-data for a data warehouse?
Yes, the meta-data (typically held in a data dictionary) contains information about the data stored in the data warehouse. It may include such information as where the data were extracted from and what they mean. This could prove to be very helpful if there are data errors or misunderstandings on the meaning of specific fields (i.e., sales date could mean date a contract was signed with a customer, or when the customer paid for the item, or the day on which the customer received the item).
What is a data mart and why is it used?
A data mart is a local single-subject database. It is often used in a tiered or federated distributed database. It is used to create a smaller database than the data warehouse for the specific needs of a group of analysts.
4.
If a company plans to include video clips in their data warehouse, what type of database would they need to create?
A relational database can store nontraditional data types (such as video clips) using BLOB or CLOB.
5.
Imagine you work as an analyst for a major mutual fund. What techniques would you use for the following questions (one word or phrase answers are appropriate):
| a | What were last year’s sales of utility funds by quarter and sales office? | OLAP |
| b | How should the market for growth funds be segmented? | Clustering |
| c | What is the potential revenue stream from a client who invests her retirement funds with the fund? | Prediction |
| d | What is common about investors in tax-exempt funds? | Classifying |
| e | What new funds do customers invest in during the twelve months after investing in their first mutual fund? | Sequential patterns |
| f | How do bond fund investors differ from other customers? | Classifying |
| g | Assisting brokers to recommend funds to prospective investors. | Decision tree |
| h | Identifying the characteristics of investors who invest in growth and index funds. | Association |
6.
You are designing a data warehouse for a hospital that wants integrated information about insurance companies (type: regional or national)), about its care facilities (location), doctors (specialty), patients (illness) and time period (week, month, quarter, year). It wants to determine the amount billed to and received from the insurance company, and the amount billed to and received from the patient. These data should be analyzable by doctor, patient, and care facility. Create the star schema.
| This page is part of the promotional and support
material for Data Management (open edition) by Richard T. Watson For questions and comments please contact the author |