We have just seen that the "Best Foot Forward" company wants to follow the progress of its shoe sales, for example, by month, style and outlet. In OLAP terminology these analysis criteria are dimensions, sometimes called axes. 数据挖掘研究院
We can, for example, talk about the Outlet dimension. There can be several values here, for example "Paris Bastille". These values are positions in the Outlet dimension.
With reference to the relational model, dimensions therefore correspond to the branches of the star schema we saw in stage 1. There is, however, a small exception: generally we would not define a month table in a relational model, because the names and orders of the months are implicit. But here we absolutely need a month dimension, which will have positions "January 2000" etc.
In general, a dimension has from two to a maximum of several thousand positions. 数据挖掘研究院
For the present, the two indicators followed by the company′s management are the quantity of shoes sold and the pre-tax sales. These indicators are called measures, or sometimes variables. 数据挖掘研究院
We can now say that the measure Quantity has the dimensions Month, Style and Outlet. 数据挖掘研究院
Each measure may have from one to several million of values. All the values of a measure are of the same data type, for example integer or decimal.
There we are! This time there is no doubt that our database has truly become multidimensional. Fortunately the measure "Quantity" only has three dimensions. It is thus still relatively feasible to represent graphically on a two dimensional screen that it really looks like a cube. The 850 shoes from the preceding page are thus displayed like this: 数据挖掘研究院
Each tile of the cube represents a value. The dimensions are indicated on the outlines of the cube. Each face of the cube corresponds to all the values for a single position of one of the three dimensions. For example, the front face is that of the Paris Bastille Outlet. Finally, the complete cube represents a measure, here the quantity of shoes sold. 数据挖掘研究院
When adding the Total Value TE, we obtain a new object with 4 dimensions: Month, Style, Outlet and Indicator, whose positions are existing measures. This new object is called hypercube.
