DMX - SQL SERVER 数据挖掘聚类

SQL SERVER 数据挖掘聚类算法对比分析

最新推荐文章于 2024-10-03 21:30:00 发布

原创

最新推荐文章于 2024-10-03 21:30:00 发布 · 2.2k 阅读

2 ·

CC 4.0 BY-SA版权

本文探讨了SQL SERVER数据挖掘中的聚类算法，包括SCALEABLE EM、NO SCALEABLE EM、SCALEABLE KM和NO SCALEABLE KM四种，并通过案例对比了它们的性能。根据测试结果，NO SCALEABLE EM算法在Case Likelihood指标上表现最优。

聚类有如下特征：

如果列标为，

1，PREDICT，预测可选；

2， INPUT，预测不可选；

3，PREDICT ONLY，TRAINING忽略。

一共有四种算法：

1，SCALEABLE EM

2， NO SCALEABLE EM

3， SCALEABLE KM

4， NO SCALEABLE KM

下面一个例子比较四种算法

create mining structure [Clustering Method]
(
[Age] long discretized(automatic,10),
[Bike Buyer] long discrete,
[Commute Distance] text discrete,
[Customer Key] long key,
[Education] text discrete,
[Gender] text discrete,
[House Owner Flag] text discrete,
[Marital Status] text discrete,
[Number Cars Owned] long discrete,
[Number Children At Home] long discrete,
[Occupation] text discrete,
[Region] text discrete,
[Total Children] long discrete,
[Yearly Income] double continuous
)

alter mining structure [Clustering Method]
add mining model [Clutering_SEM]
using microsoft_clustering
(CLUSTERING_METHOD = 1)

alter mining structure [Clustering Method]
add mining model [Clutering_NSEM]
using mic