A versatile approach to classification of multivariate time series data

During the recent decade we have experienced a rise of popularity of sensors capable of collecting large amounts of data. One of most popular types of data collected by sensors is time series composed of sequences of measurements taken over time. With low cost of individual sensors, multivariate time series data sets are becoming common. Examples can include vehicle or machinery monitoring, sensors from smartphones or sensor suites installed on a human body. This paper describes a generic method that can be applied to arbitrary set of multivariate time series data in order to perform classification or regression tasks. This method was applied to the 2015 AAIA Data Mining Competition concerned with classifying firefighter activities and consecutively led to achieving the second-high score of nearly 80 participant teams.

[1]  Dominik Slezak,et al.  Tagging Firefighter Activities at the emergency scene: Summary of AAIA'15 data mining competition at knowledge pit , 2015, 2015 Federated Conference on Computer Science and Information Systems (FedCSIS).

[2]  Ian H. Witten,et al.  The WEKA data mining software: an update , 2009, SKDD.

[3]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[4]  Dominik Slezak,et al.  Key risk factors for Polish State Fire Service: A Data Mining Competition at Knowledge Pit , 2014, 2014 Federated Conference on Computer Science and Information Systems.