column-distributions-data-mining.md

title	Column Distributions (Data Mining) | Microsoft Docs
ms.custom
ms.date	06/13/2017
ms.prod	sql-server-2014
ms.reviewer
ms.technology	analysis-services
ms.topic	conceptual
helpviewer_keywords	normal distribution type [data mining] uniform distribution type [data mining] columns [data mining], distributions log normal distribution type [data mining] continuous columns distributions [data mining]
ms.assetid	87e700de-32be-4bc8-b01d-ba4ee1ab48de
author	minewiskan
ms.author	owend
manager	craigg

Column Distributions (Data Mining)

In [!INCLUDEmsCoName] [!INCLUDEssNoVersion] [!INCLUDEssASnoversion], you can define column distributions in a mining structure, to affect how algorithms process the data in those columns when you create mining models. For some algorithms, it is useful to define the distribution of any continuous columns before you process the model, if the columns are known to contain common distributions of values. If you do not define the distributions, the resulting mining models may produce less accurate predictions than if the distributions were defined, because the algorithms will have less information from which to interpret the data.

The algorithms that are available in [!INCLUDEssASnoversion] support the following distribution types:

Normal The values for the continuous column form a histogram with a normal distribution.

Log Normal The values for the continuous column form a histogram, where the curve is elongated at the upper end and is skewed toward the lower end.

Uniform The values for the continuous column form a flat curve, in which all values are equally likely.

For more information about the algorithms that [!INCLUDEssASnoversion] provides, see Data Mining Algorithms (Analysis Services - Data Mining).

See Also

Content Types (Data Mining) Mining Structures (Analysis Services - Data Mining) Discretization Methods (Data Mining) Distributions (DMX) Mining Structure Columns