HTM is well suited for applications that have the following characteristics:
- Data flowing through time: the data can be in the form of numbers, dates, text, or GPS points
- A data sampling rate from once per minute to once per hour, with the “sweet spot” being between once per minute and once every five minutes (faster velocity data can be aggregated or sampled as well)
- Data that has inherent structure, i.e. not entirely random
- Many models are required rather than one large model
- Focus of the application is prediction or anomaly detection
The following applications are examples that fit these characteristics:
- Highlighting anomalies in the behavior of moving objects, such as tracking a fleet’s movements on a truck by truck basis
- Understanding if human behavior is normal or abnormal on a securities trading floor
- Predicting energy usage for a utility on a customer by customer basis
- Predicting failure in a complex machine based on data from many sensors
In order to demonstrate these capabilities, we have created example applications.
Rogue Behavior Detection
This example application models normal behavior of individuals by detecting changes in behavior, such as unusual use of files. You can experiment with this application using your own data by downloading our sample application code below.
The geospatial tracking application detects anomalies in the movement of people, objects, or material using speed and location data. Use this application to enable logistics optimization. You can experiment with this application using your own data by downloading our sample application code below.
HTM for Stocks is an example application that detects anomalies in publicly traded companies. It continuously models stock price, stock volume, and Twitter volume and alerts you when something unusual is happening. You can experiment with this application by downloading our sample application code below and connecting it to Twitter and metric collectors.