A Novel Technique to Time Series Shapelets

Full report available here

 

Time Series Shapelets

 

Setting:

  1. A pool of time series, of different length

  2. They are from class A and B, labeled

 

Goal:

       Find a series of shapelets to optimally split the set

 

Idea:

  1. Supervised learning (explicit labeling), training / testing

  2. Decision tree (find an optimal shapelet at each node)

  3. Euclidean distance (basically distance measure only computed for subsequences with the same length)

  4. Some modification applied to brute force method to reduce complexity and storage

 

 

Definition:

 

 

 

 

 

 

 

Brute force algorithm:

 

 

 

 

Modification:

 

1.