The SELECT statement in Data Mining Extensions (DMX) is used for the following tasks in data mining:
Browsing the content of an existing mining model
Creating predictions from an existing mining model
Creating a copy of an existing mining model
Browsing the mining structure
Although the full syntax of this statement is complex, the primary clauses used for browsing a model and its underlying structure can be summarized as follows:
SELECT [FLATTENED] [TOP <n>] <select list> FROM <model/structure>[.aspect] [WHERE <condition expression>] [ORDER BY <expression>[DESC|ASC]]
Some data mining clients cannot accept result sets in hierarchical format from a data mining provider. The client may lack the ability to handle a hierarchy, or it may have to store the results in a single denormalized table. To convert the data from nested tables to flattened tables, you must request that the query results be flattened.
To flatten the query results, use the SELECT syntax with the FLATTENED option, as shown in the following example:
SELECT FLATTENED <select list> FROM ...
You can order the results of a query by using an expression, and can then return a subset of the results by using a combination of the ORDER BY and TOP clauses. This is useful in a scenario such as targeted mailing where you only want to send results to the most likely respondents. You could order the results of a target mailing prediction query by the prediction probability, and then only return the top <n> results.
The <select list> can include scalar column references, prediction functions, and expressions. The options that are available depend on the algorithm, and the following contexts:
Whether you are querying a mining structure or a mining model
Whether you are querying content or cases
Whether source data is a relational table or a cube
If you are making predictions
In many cases, you can use aliases, or create simple expressions based on the items in the select list. For example, the following example shows a simple expression on model columns:
SELECT [CustomerID], [Last Name] + ', ' + [FirstName] AS FullName FROM <model>.CASES
The following example creates an alias for a column that contains the results of a prediction function:
SELECT Predict([Column1], 'Value') as Column1Prediction FROM MyModel JOIN <source data query>
You can limit the cases that are returned by the query by using a WHERE clause. The WHERE clause specifies that column references in the WHERE expression must have the same semantics as column references in the <select list> of the SELECT statement, and can only return a Boolean expression. The syntax for the WHERE clause is as follows
WHERE < condition expression >
The select list and WHERE clause of a SELECT statement must follow the following rules:
The select list must contain an expression that does not return a Boolean result. You can modify the expression, but the expression must return non-Boolean results.
The WHERE clause must contain an expression that returns a Boolean result. You can modify the clause, but it must return a Boolean result.
There are two types of syntax that you can use for creating predictions:
The first type of prediction enables you create complex predictions either in real time or as a batch.
The second prediction type creates an empty prediction join on a predictable column in a mining model, and returns the most likely state of the column. The results of this query are completely based on the content of the mining model.
You can insert a select statement into the source query of a SELECT FROM PREDICTION JOIN statement by using the following syntax.
SELECT FROM PREDICTION JOIN (<SELECT statement>) AS t, WHERE <SELECT statement>
For more information about creating prediction queries, see Prediction Queries (DMX).
Because of the complexity of browsing with the SELECT statement, detailed syntax elements and arguments are described by clause. For more information about each clause, click a topic in the following list: