Data Mining Algorithms – Analysis Services

Supported Data Mining Algorithms

Analysis Services provides a variety of algorithms for classification, clustering, and association tasks. Choose the algorithm that best fits your data characteristics and business goals.

Classification

Clustering

Association

Classification Algorithms

Decision Trees (Microsoft Decision Trees) – intuitive, fast, and works well with mixed data types.
Neural Networks – suitable for complex, non‑linear relationships.
Support Vector Machines (Linear) – high‑dimensional data with clear margins.
Logistic Regression – binary outcomes with probability estimates.
Naïve Bayes – efficient for large text datasets.

CREATE MINING MODEL [CustomerChurn]
  FROM [AdventureWorksDW2019].[dbo].[DimCustomer]
  USING Microsoft_Decision_Trees
  (
    TargetColumn = [ChurnFlag],
    PredictContinuousColumn = TRUE
  );

Clustering Algorithms

K-Means – fast, works well with numeric data.
Self‑Organizing Maps (SOM) – captures non‑linear structures.
Hierarchical Clustering – builds a tree of clusters.

CREATE MINING MODEL [ProductSegmentation]
  FROM [AdventureWorksDW2019].[dbo].[FactResellerSales]
  USING Microsoft_KMeans
  (
    K = 5,
    TrainingSet = N'Top 1000'
  );

Association Algorithms

Microsoft Association Rules – discovers frequent itemsets and rules.

CREATE MINING MODEL [MarketBasketAnalysis]
  FROM [AdventureWorksDW2019].[dbo].[FactInternetSales]
  USING Microsoft_Association_Rules
  (
    Support = 0.01,
    Confidence = 0.6
  );

Choosing the Right Algorithm

Consider the following factors when selecting an algorithm:

Scenario	Recommended Algorithm
Predicting churn (binary outcome)	Decision Trees or Logistic Regression
Segmenting customers	K-Means or SOM
Finding product bundles	Association Rules

Interactive Demo

Select a dataset and view a preview of model predictions.

Dataset: