ShapeletUnsupervised Example - Teradata Vantage

Input

The input table has 10 price observations for four stocks. The column period contains time values, represented by integers. Because the function is unsupervised, it ignores the stock_category column; however, you can use that column to verify that the created clusters belong to the same category.

The following figure is a graphic representation of the input data.

In the time period shown, technology stocks 1 and 4 have similar price trajectories, as do healthcare stocks 2 and 3.

SQL Call

SELECT * FROM ShapeletUnsupervised (
	ON ushapelets_input AS InputTable
	OUT TABLE OutputTable (uss_output)
	USING
	IDColumn ('stockid')
	TimeColumn ('period')
	TargetColumn ('stockprice')
	SAXWindowSize (5)
) AS dt;

Output

 stats                                 
 ------------------------------------- 
 Unsupervised shapelets table created.
 number of clusters : 2               
 number of timeseries : 4

The function assigned technology stocks 1 and 4 to cluster 0 and healthcare stocks 2 and 3 to cluster 1.

SELECT * FROM uss_output;

 stockid cluster_label 
 ------- ------------- 
       4             0
       1             0
       2             1
       3             1

Download a zip file of all examples and a SQL script file that creates their input tables from the attachment in the left sidebar.