Summary
Overview
Anomaly detection is a way of detecting abnormal behavior. This template uses an autoencoder machine learning model or LSTM method to specify expected behavior and then monitors new data to match and highlight unexpected behavior. Version 2 features automated machine learning to optimize model tuning parameters. The Time Series releases include time series analysis, so it can be used as a form of 'control chart', as well input component drill-down to find the most important features influencing a reconstruction error and clustering analysis to group and analyze similar groups of anomalies.
Details
You can try this template online on our demo gallery.
If you want to try the template on your data, feel free to download the template.
More details about this important template can be found in this community article or in the Dr. Spotfire recording below (in this recording older version 4.0.0 is shown).
Release 5.0.1 for Time Series Analysis
Published: October 2023
Changes made to previous version:
- Extending the documentation with details about the data preparation functions
- Bug fix in the script tile_to_cutoff.py
Release Includes:
- Template dashboard
- Documentation
- License information
- Python custom modules used by the template
Release 5.0.0 for Time Series Analysis
Published: August 2023
Changes made to previous version:
- Possibility to train more models at once
- Python class architecture to make template easily extensible
- Adding LSTM method as modelling method
- Simplifying changing the data source and overall user experience
Release Includes:
- Template dashboard
- Documentation
- License information
- Python custom modules used by the template
Release 4.1.0 for Time Series Analysis
Published: January 2023
Changes made to previous version:
- Ensuring compatibility with Spotfire version 12.0 as well as recent TensorFlow package
- Improvements to scalability, performance, and handing of missing data and edge cases
Release Includes:
- Template dashboard
- Documentation
- License information
Release 4.0.0 for Time Series Analysis
Published: November 2021
Changes made to previous version:
- Autoencoder functionality is now done through Python data function (if you prefer H2O implementation, please download version 3.0 for Time Series Analysis)
- Major look and feel enhancements
Release Includes:
- Template dashboard
- Documentation
- License information
- Network Chart Mod v1.0.1 installer (this visualization is used inside the template)
Package compatibility for version 4.0.0:
This analysis has been tested with Spotfire 11.4
- Python packages used (pandas-1.2.3, numpy-1.19.5, scipy-1.6.3, scikit-learn-0.24.2, tensorflow-2.9.1)
- TERR: CRAN packages used (data.table-1.13.0, lubridate-1.7.9, SpotfireUtils)
Release 3.0.0 for Time Series Analysis
Published: January 2019
Minor UI changes
This analysis has been tested with Spotfire 7.10, TERR 4.4.0, CRAN packages data.table (version 1.10.4)
and h2o (version 3.20.0.2). More info about this release in this video.
Release 2.0.0 for Time Series Analysis
Published: January 2019
This release contains bug fixes and added function to specify the time units on the summarize to day page. Also, all functionality in the original P1.0 Time Series release to enable time series analysis and clustering of anomalies
Release 2.0.0
Published: January 2018
Addition of option to automate optimization of of model tuning parameters. This facilitates use by a business user or citizen data scientis
Release 1.0.0 for Time Series Analysis
Published: December 2018
This release contains additional functionality to enable time series analysis and clustering of anomalies
Release 1.0.0
Published: March 2017
Initial Release