复旦商务智能概论--数据仓库DataWarehouse-公开课件（讲义）.ppt

* The model of a data warehouse is similar to the model used for Online Transactional Processing (OLTP) data, but has some very important differences. The difference includes aspects such as historical/temporal nature of the data, heavy use of summarizations and aggregations, and accessing data in a total business subject area rather than a specific business process. The warehouse user will employ many interactive queries and analysis techniques against high volumes of data instead of the predefined, controlled set of functions used for OLTP applications. * The star schema is recommended while developing the logical data model, and then converting or transforming it to whatever physical structure will be used for the implementation of the model. A normalized snowflake schema is very similar to the entity relationship model. * Notice that each of the dimension entities has a unique identifier and these identifier attributes are also in the fact entity. This is the connector that supports the relationships between each dimension entity and the fact entity. * The logical model is translated into an initial physical model for the database management system to be used for implementation. For example, in a relational implementation, each entity in the model is identified as a table in the relational database. * Granularity-The lowest level of detail for the measures in a fact; the combination of the lowest level of detail in each dimension of the fact. * Dimension-A collection of business data which is related to and gives meaning or relevance to the measure. Fact-A collection of measures and the unique identifiers of their related dimensions. Measure- A numerically valued quantifier of a business activity or item. * Aggregation level - A set of attributes that qualify a desired summarization of measure values like sum of sales revenue by city and customer for the last 6 months. * Dimension Hierarchy- The structure of a dimension containing base or gra