Skip to content
ISETS: Incremental Shapelet Extraction from Streaming Time Series
Python HTML Shell Makefile
Branch: master
Clone or download

Latest commit

Fetching latest commit…
Cannot retrieve the latest commit at this time.

Files

Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
Datasets ISETS May 15, 2019
ISETS_webapp
Spark_Cluster_Docker
ISETS.py ISETS May 15, 2019
PKDD_ISETS_video.mp4 Video of ISETS May 15, 2019
README.md Update README.md Dec 4, 2019
SMAP_block.py ISETS May 15, 2019
evaluation_block.py ISETS May 15, 2019
memory_block.py ISETS May 15, 2019

README.md

ISETS

Incremental Shapelet Extraction from Streaming Time Series

-Jingwei ZUO, Karine ZEITOUNI, Yehia TAHER

DAVID Lab, University of Versailles Saint-Quentin, Unverisy of Paris-Sacaly

This web application is intended to provide users an intuitive understanding about the feature extraction process in a combined context of Time Series and Data Streams. With ISETS, users can monitor the occurence of Concept Drift and the Shapelet Ranking at different Time Points.

Project Page: Incremental and Adaptive Feature Exploration over Time Series Stream

Demo 1 -> ISETS Tutorial

Demo 2 ->ISETS and Adaptive Features

Configurations (demo)

  • Input File: the name should be end with "Train.csv"
  • dataset_folder: in each file, change the location of the datasets in the background. The selected input file will be saved/uploaded into this folder.
  • Data Augmentation: refer to preprocessing/TS_stream_preprocess.py. As Shapelet-based methods (e.g., SMAP) are noise resistant, we put randomly the noise of random durations into the original TS data to augment the data volume.

####Web application (demo)

  • ISETS_webapp.py: main program, a web application based on Flask and Bokeh
  • ISETS_webbackend.py: the program for adaptive shapelet extraction and Concept Drift detection
  • draw_adaptive_shapelets.py: show the adaptive shapelets in the web interface
  • draw_TS_Stream.py: show in real time the input TS instances in the stream

###Core algorithms

  • utils/: the repository which contains the basic file operations and similairty measure functions
  • memory_block.py: the caching mechanism including the computation of Matrix Profile for cached instance
  • SMAP_block.py: Shapelet extraction on MAtrix Profile
  • evaluation_block.py: the loss computation and the Concept Drift detection on TS Stream
  • adaptive_features/adaptive_features.py: Concept Drift detection and adaptive feature extraction
  • ISMAP/ISAMP.py: incremental Shapelet extraction on MAtrix Profile
You can’t perform that action at this time.