Supervised learning with multi-scale time intervals using a statistical classification model to classify unlabeled events

A computer-readable storage medium encoded with program code for generating a statistical classification model used by a computer system to determine a class associated with an unlabeled time series event, the program code comprising:

• program code for receiving a set of labeled time series events, wherein each time series event is labeled with a class label;
• program code for identifying, for each time series event of the set of labeled time series events, a set of time series features, wherein each time series feature represents a feature for at least one time point of a time series event;
• program code for identifying, for each time series event of the set of labeled time series events, a plurality of time intervals associated with at least a first time series feature of the set of time series features, wherein each time interval includes a plurality of time points associated with feature values above a specified threshold of a plurality of specified thresholds associated with the at least a first time series feature;
• program code for generating, for each time series event of the set of labeled time series events, a plurality of multi-scale features based on the plurality of time intervals;
• program code for identifying, for each of at least some time series events of the set of labeled time series events, a first subset of the plurality of multi-scale features that correspond at least in part to a subset of space or time points within a time series event that contain feature data that distinguish the time series event as belonging to a class of time series events that corresponds to the class label;
• program code for generating a statistical classification model for classifying an unlabeled time series event based on the class corresponding with the class label, based at least in part on the at the first subset of the plurality of multi-scale features generated for the at least some time series events of the set of labeled time series events;
• and
• program code for storing the statistical classification model in a computer-readable storage medium.

A computer-implemented system, for generating a statistical classification model used by a computer system to determine a class associated with an unlabeled time series event, the system comprising:

• a database for storing a set of labeled time series events, wherein each time series event is labeled with a class label;
• a time series feature generation module adapted to identify, for each time series event of a set of labeled time series events, a set of time series features, wherein each time series feature represents a feature for at least one time point of the time series events;
• a multi-scale feature generation module adapted to:
• identify, for each time series event of the set of labeled time series events, a plurality of time intervals associated with at least a first time series feature of the set of time series features, wherein each time interval includes a plurality of time points associated with feature values above a specified threshold of a plurality of specified thresholds associated with the at least a first time series feature;
• and
• generate, for each time series event of the set of unlabelled time series events, a second plurality of multi-scale features based on the plurality of time intervals;
• and
• a classifier construction engine adapted to:
• generate a statistical classification model for classifying an unlabeled time series event based on the class corresponding with the class label, based at least in part on the at the first subset of the plurality of multi-scale features generated for the at least some time series events of the set of labeled time series events;
• and
• store the statistical classification model in a computer-readable storage medium.