What is a Multi-dimensional Model?

Multi-dimensional model is an integral aspect of the On-line Analytical Processing which also known as OLAP.

Due to the fact that OLAP is online it provides information quickly, iterative queries are often posed during interactive sessions.

Due to the analytical nature of OLAP the queries are often complex. The multi-dimensional model is used to solve this kind of complex queries. The model is important because it applies simplicity.

This helps users understand the databases and enables software to plot a course through the databases effectively.

Multi-dimensional data models are made up of logical cubes, measures, and dimensions. Within the models you can also find hierarchies, levels, and attributes.

The straightforwardness of the model is essential due to the fact that is identifies objects that represent real world entities.

The analysts know what measures they want to see, what dimensions and attributes make the data important, and in what ways the dimensions of their work is organized into levels as well as hierarchies.

What are Logical Cubes and Logical Measures?

Let us touch on what the logical cubes and logical measures are before we move on to more complicated details.

Logical cubes are designed to organize measures that have the same exact dimensions. Measures that are in the same cube have the same relationship to other logical objects; they can easily be analyzed and shown together.

With logical measures cells of the logical cube are filled with facts collected about an organization’s operations or functions.

The measures are organized according to the dimensions, which also deals with time dimension.

Analytic databases contain outlines of historical data, taken from data in a heritage system, also those other data sources such as syndicated sources. The normally acceptable amount of historical data for analytic applications is about three years worth.

The measures are static; they are also trusted to be consistent while they are being used to help make informed decisions.

The measures are updated often, most applications update data by adding to the dimensions of a measure. These updates give users concrete historical record of a specific organizational activity for an interval. This is very productive.

Another productive strategy is that adopted by other application, which fully rebuild the data rather then perform updated.

The lowest level of a measure is called the grain. Often this level of data is never seen, even so it has a direct affect on the type of analysis that can be done.

This level also determines whether or not the analysts can obtain answers. Questions such as, when are men most likely prone to place orders for custom purchases?

{qbapagebreak title=Logical Dimensions}

Logical Attributes, Dimensions, Hierarchies and Levels 

Logical cubes and measures were relatively simple and easy to digest. Now we will consider Logical Dimensions, which is a little more complex. Dimensions have a unique set of values that define and categorized the data.

These form the sides of the logical cubes and through this the measures inside of the cubes as well. The measures themselves are usually multi-dimensional; due to this a value within a measure should be qualified by a member of all of the dimensions in order to be appropriate.

The Hierarchy is a mode which is used to organize the data at each level of aggregation. When looking at data, developers use hierarchy dimensions to identify trends on a specific level, as well as drill down to lower lever to indicate what is causing such trends, then they can also roll up to the higher levels to view how these trends affect the bigger sections of the organization.

Back to the levels, each level represent a position in the hierarchy, the levels above the most detailed level contain aggregated values for the levels that are beneath it.

On different levels, the members of those levels have a hierarchical relation, which we defined in the article related to this topic as a parent / child relationship, where the parent can have many children but there can only be one parent to a child.

The hierarchies and levels have a many-to-many relationship, the hierarchy is usually consisted of many levels, and one level can be includes in various hierarchies.

Finally to wrap up this section we will take a quick look at Logical Attributes. By now we should all know that an attribute provides extra information about the data.

Some of there attributes are used simply for display. You can have attributes that are like, flavors, colors, sizes, the possibilities are endless.

It is this kind of attribute that can be helpful in data selection and also in answering questions.

An example of the type of questions that the attributes can help answer are; what colors are most popular in abstract painting? Also we can ask, what flavor of ice-cream do seven year olds prefer?

We also have time attributes, which can give us information about the time dimensions we spoke of earlier, this information can be helpful in some kinds of analysis.

These types of analysis can be indication the last day or amount of days in a time period. That pretty much wraps it up for attributes at this point. We will revisit the topic a little later.

Variables

Now we will consider the issue of variables. A variable is basically a value table for data, which is an array with a specific type of data and is indexed by a particular list of dimensions. Please be sure to understand that the dimensions are not stored in the variable.

Each mixture of members of a dimension define a data cell. This is true whether a value for that cell is present or not. Therefore, if data is missing, or absent the fact of the absences can either be included or excluded from analysis.

There is no specific relationship between variables that share like dimensions. Even so, a logical relationship does exist between then, this is due to the fact that even though they may store different data that could be from a different data type, they are identical containers.

When you have variable that contain identical dimensions it creates a logical cube. With that in mind, you can see how if you change a dimension, like adding time periods to the time dimension then the variables change as well to include the new time periods, this happens even if the other variable have no data for them.

The variables that share dimensions can be manipulated in a array of ways, this includes aggregation, allocation, modeling, and calculations.

This is more specifically numeric calculations, and it is an easy and fast method in the analytic work place.  We can also use variables to store measures.

In an analytic work place factual information is kept in variables, normally they are kept with a numeric data type.

Each data type is then stored in an associated variable, this is so that while sales and expense data may have like dimensions and the same data type, they will be stored in distinct variables.

In addition to using variable to store measures they can be used to store attribute as well. There are major differences between the two.

While attribute are multi-dimensional, only one dimension is the data dimension. Attributes give us information about each dimension member no matter what level it inhabits.

Through out our journey of learning about the different types of data models, I think that the multi-dimensional model is perhaps one of the most useful.

It takes key aspects from other models like the relational mode, the hierarchical model, and the object model, and combines those aspects into one competent database that has a wide variety of possible uses.