Algorithm Reference (Analysis Services - Data Mining)
This section provides links to topics that provide additional information about specific data mining algorithms. This section also provides a list of the functions that can be used with each algorithm.
For an overview of how data mining algorithms work, or the various business scenarios where you would benefit from using a particular algorithm, see Data Mining Algorithms (Analysis Services - Data Mining).
General Description of Algorithm
Choosing the right algorithm for the analytic task, and preparing data to meet the requirements of analysis, are important steps in the data mining process. The following topics provide an overview of how each algorithm works, sets out an example of an analytic task for which the algorithm is suited, and describes how the model is used in the scenario. Each topic also contains a Requirements section that provides guidelines on the type of data that is required for each model type.
Algorithm Technical Reference
When you select an algorithm to use in creating a model, you can accept the defaults provided by Analysis Services, but in many cases you might need to customize the way the model is created or the way the algorithm processes data. The following topics describe the parameters that you can use to customize your mining model, and also provide detailed technical information about the implementation of each algorithm.
When you build a model, you can customize the model and potentially affect the results by filtering the data that is used when training the model. For more information about how to use filters when training and testing mining models, see Creating Filters for Mining Models (Analysis Services - Data Mining) and Tools for Charting Model Accuracy (Analysis Services - Data Mining).
Query and Prediction Function Reference
You can use functions to retrieve the results of a mining model. A prediction function can provide detailed information about patterns and statistics found in analysis, or it can be used for making predictions and filtering predictions based on probability or importance.
For a list of all prediction functions, see Data Mining Extensions (DMX) Function Reference.
The following table lists the functions in Analysis Services that can be used for creating queries on all algorithm types.
Using Prediction Functions with Specific Model Types
Because each algorithm creates different patterns, there are additional prediction functions that are unique to the each model type. The way that the prediction functions are used, and the way the results are interpreted, might also change slightly depending on the mining model. For examples of how to use prediction functions to create queries on specific model types, see the following topics: