Multidimensional mode

The Multidimensional mode uses the analytical multidimensional cubes for analyzing data across multiple dimensions. Once you create a new database in Analysis Services, you can see the different features displayed inside the database, in the same way as you can in Data Tools while developing the Analysis Services project:

The Data Source and Data Source Views will define what data you work with in the Analysis Services project. In most cases, this will be a database, or a data warehouse database, created by Integration Services to provide proper aggregations, denormalization of the data, and interconnection between various different information systems.

Once you have defined the Data Source Views, you can start building your multidimensional model by creating dimensions and cubes. Dimensions are objects used to get information about fact data in a cube or more cubes. Common attributes for date dimensions are year, quarter, week, month, and so on. For product dimensions, the common attributes might include name, category, size, price, and so on. When working with such attributes, you can usually build hierarchies based, so first you would see data aggregated by year, quarter, month, and then week if you drill down.

Once you have created all the objects which you need, you can deploy the project to your Analysis Services service and access the data via either Management Studio or other tools that can be used for data analysis:

Important features used for data science in Analysis Services are cubes, dimensions, and mining models. Once the cube and dimensions are created, you can create several mining models to get a larger insight into the data that is processed on SSAS. There are several mining models available, depending on the nature of the data and the task being accomplished. You can choose from the following:

• Association rules
• Clustering
• Decision trees
• Linear regression
• Logistic regression
• Naive Bayes
• Neural networks
• Sequence clustering
• Time series

Here is a screenshot of the available mining models:

These mining models get data from the mining structure and provide data analysis based on the specific algorithm. Until the structure has been processed and analyzed, the mining model will be empty. Once the model has been processed, you can see the results via Management Studio, or SQL Server Data Tools.

If we consider all the available algorithms, they can be split into the following categories:

• Regression algorithms predict numeric values, such as profit or loss 
• Classification algorithms predict discrete variables 
• Segmentation algorithms group data into clusters or find groups of items with similar properties
• Sequence analysis algorithms can be used for finding series, such as log events in the server log
• Association algorithms find correlation between attributes