As the number of sensors increases due to the development of Internet of Things (body sensors, weather station, ...) and the improvement of existing devices (satellite, chirurgical robot, ...), the number of available time series data increases also and so is their temporal sampling in a variety of applications in medicine (e.g., EEG and ECG signals), environment, finance, weather forecasting, food security, or human activity recognition. There is now a need to deal with new issues such as missing values, irregular temporal sampling, time series outliers, early classification, heterogeneous time series, cold start and the lack (or few amount) of labeled data to train models. Moreover, new devices with higher temporal frequency and multi sensors impose to set a tradeoff between accuracy and scalability as the dimensionality and the length of the time-series produced by these devices increase. Finally, these multivariate time series require new methods to deal with the dimensionality complexity. The aim of this special session is to present recent work on time series analysis in different application domains, that deals with the increase in time series data and/or in the temporal sampling.
TOPICS
Time-Series classification
Time-Series forecasting
Time-Series clustering
Time-Series segmentation
Time-Series regression
Time series at scale
Early classification of Time-Series
Time-Series classification with few labeled data
Time-Series classification with missing data and/or outliers
PAPER SUBMISSION
Authors are invited to submit papers in any of the topics listed above.