Integrated Computational Materials, Process and Product Engineering – IC-MPPE 2022

Searching, finding and annotating patterns in time series data has become even more important with the advent of machine learning systems, which rely on massive amounts of training data to analyse, model and predict system behaviour. We present TimeFuse, a set of algorithms integrated in a software suite which enables engineers to find patterns in time series data based on selected similar data or sketched signal shapes. TimeFuse facilitates rapid identification and annotation of similar patterns in time series data. We employ an information retrieval approach based on the SAX algorithm to identify signal ranges similar to a specified search pattern with a strong emphasis on recall over precision. A streamlined user interface enables users to rapidly select, group and annotate identified clusters of similar signals. With TimeFuse, users can quickly prepare training data for modelling purposes based on sketches of their understanding of expected signal shapes or based on known patterns in historical data.