Ter this article
APPLIES TO: SQL Server Analysis Services Azure Analysis Services
Gegevens mining queries are useful for many purposes. You can:
Apply the proefje to fresh gegevens, to make single or numerous predictions. You can provide input values spil parameters, or ter a batch.
Get a statistical summary of the gegevens used for training.
Samenvatting patterns and rules, or generate a profile of the typical case indicating a pattern ter the specimen.
Samenvatting regression formulas and other calculations that explain patterns.
Get the cases that gezond a particular pattern.
Retrieve details about individual cases used te the monster, including gegevens not used te analysis.
Retrain a proefje by adding fresh gegevens, or perform cross-prediction.
This section provides an overview of the information you need to get embarked with gegevens mining queries. It describes the types of queries you can create against gegevens mining objects, introduces the query devices and query languages, and provides linksaf to examples of queries that you can create against models that were built using the algorithms provided ter SQL Server Gegevens Mining.
Understanding Gegevens Mining Queries
Analysis Services Gegevens Mining supports the following types of queries:
Queries that make inferences based on patterns ter the proefje, and from input gegevens.
Queries that comeback metadata, statistics, and other information about the specimen itself.
Queries that can retrieve the underlying case gegevens for the specimen, or even gegevens from the structure that wasgoed not used te the monster.
Queries that do not come back information from the prototype, but rather are used to build models and structures or to update the gegevens ter a prototype or structure.
Before you create queries, wij recommend that you familiarize yourself with the differences inbetween models created with each of the gegevens mining algorithms provided by SQL Server.
Browse and explore each monster type by using the custom-built gegevens mining viewers that are provided for each algorithm type. For more information, see Mining Specimen Viewer Tasks and How-tos.
Review the proefje content for each monster type, by using the Microsoft Generic Content Tree Viewer. To interpret this information, refer to Mining Proefje Content (Analysis Services – Gegevens Mining),.
Query Contraptions and Interfaces
You can build gegevens mining queries interactively by using one of the query instruments provided by SQL Server. The graphical Prediction Query Builder is provided te both SQL Server Gegevens Contraptions (SSDT) and SQL Server Management Studio. If you have not used the Prediction Query Builder before, wij recommend that you go after the steps ter the Basic Gegevens Mining Tutorial to familiarize yourself with the interface. For q quick overview of the steps, see Create a Query using the Create a Prediction Query Using the Prediction Query Builder.
The Prediction Query Builder is helpful for embarking queries that you will customize zometeen. You can lightly add gegevens sources and opbergmap them to columns, and then switch to DMX view and customize the query by adding a WHERE clause or other functions.
Once you are familiar with gegevens mining models and how to build queries, you can also write queries directly by using Gegevens Mining Extensions (DMX). DMX is a query language that is similar to Transact-SQL, and that you can use from many different clients. DMX is the implement of choice for creating both custom-built predictions and elaborate queries. For an introduction to DMX, see Creating and Querying Gegevens Mining Models with DMX: Tutorials (Analysis Services – Gegevens Mining),.
DMX editors are provided te both SQL Server Gegevens Contraptions (SSDT) and SQL Server Management Studio. You can also use the Prediction Query Builder to begin your queries, then switch the view to the text editor and copy the DMX statement to another client. For more information, see Gegevens Mining Query Implements.
You can compose DMX statements programmatically and send them from your client to the Analysis Services server by using AMO or XMLA. However, DMX is the language that you voorwaarde use to create queries against a mining monster.
You can also query the metadata, statistics, and some content of the proefje by using Dynamic Management Views (DMVs) that are based on the gegevens mining schema rowsets. Thesis DMVs make it effortless to retrieve information about the monster by typing SELECT statements, however, you cannot create predictions. For more information about DMVs supported by Analysis Services, see Use Dynamic Management Views (DMVs), to Monitor Analysis Services.
Ultimately, you can create gegevens mining queries for use te Integration Services packages, by using the Gegevens Mining Query Task, or the Gegevens Mining Query Transformation. The control flow task supports numerous types of DMX queries, whereas the gegevens flow transformation supports only queries that work with gegevens te the gegevens flow, meaning queries that use the PREDICTION JOIN syntax.
Queries for Different Monster Types
The algorithm that wasgoed used when the prototype wasgoed created greatly influences the type of information that you can get from a gegevens mining query. The reason for the differences is that each algorithm processes the gegevens ter a different way, and stores different kinds of patterns. For example, some algorithms create clusters, others create trees. Therefore, you might need to use specialized prediction and query functions, depending on the type of proefje that you are working with.
The following list provides a summary of the functions that you can use ter queries:
General prediction functions: The Predict function is polymorphic, meaning it works with all proefje types. This function will automatically detect the type of monster you are working with and prompt you for extra parameters. For more information, see Predict (DMX),.
Not all models are used to make predictions. For example, you can create a clustering monster that does not have a predictable attribute. However, even if a prototype does not have a predictable attribute, you can create prediction queries that terugwedstrijd other types of useful information from the monster.
Custom-made prediction functions: Each proefje type provides a set of prediction functions designed for working with the patterns created by that algorithm.
For example, the Lagen function is provided for time series models, to let you view the historical gegevens used for the prototype. For clustering models, functions such spil ClusterDistance are more meaningful.
For more information about the functions that are supported for each monster type, see the following linksom: