Online-Academy
Look, Read, Understand, Apply

Data Mining And Data Warehousing

MultiDimensional Data Model

Multidimensional Data Model (MDDM)

Dimensional Modeling is a technique to represent/visualize data stored in a data Warehouse. This modeling technique helps faster retrieval of data. Ralph Kimball developed the concept of Dimensional modeling and it consists of tables known as "fact" and "dimension" to store facts and dimension records.

A multidimensional data model is designed to read, analyze and summarize numeric information like a sum of the total, a sum of expenses, balances, values, etc.
Fact :Facts are the measurements about the business like weekly sales, weekly total sales volume, monthly sales, total number of products etc.
Dimension :Dimension provides additional information like who, what, where about the fact such that fact becomes meaningful and understandable. Examples of Dimension are: Customers, suppliers, location, Product etc.
Attributes : Attributes define dimensions, that is attributes provide detail about the dimension; description, unit_price, MFD are attributes of Product dimension; year, quarter, month, week are attributes of time dimension.
Fact Table : Fact table is the biggest table in the MDDM and contains:
  • Measures/Facts
  • Foreign keys to dimension table
Dimension Table :Dimension table specifies dimensions of a fact. Dimension tables are joined to fact table with the help of foreign key.

Dimensional Modelling

Dimension models are created using following steps:
  • The business processes involved in the system of interest are identified and studied.
  • The dimensions along which the analysis should be performed are identified.
  • Level of details up to which analysis has to be performed are identified.
  • Facts/measures are identified.
  • Make connection between dimensions and the fact and form a star schema
Multidimensional Data model represents data in the form of cubes; cube is defined by dimensions and facts. Several cubes can be produced with different combinations of the considered dimensions. Those cubes form a lattice consisting of different levels of cuboids like base-cuboid, apex cuboid.
Star Schema is used to model data warehouse, there is one fact table and several associated dimension tables. The dimension tables are joined with the fact table using a foreign key. Fact table contains keys and fact/measure. Dimension tables are not normalized.
Snowflake Schema is an extension of a Star Schema, and dimension tables are normalized.
Galaxy Schema (Fact Constellation schema): contains two or more fact tables; dimension tables are shared between the fact tables.