Section: New Results

Time Series Rule Matching: Application to Energy Consumption

Participants : Maël Guillemé, Véronique Masson, Laurence Rozé, René Quiniou, Alexandre Termier.

Pattern mining in time series is an important subfield of Data Mining. In various applications, patterns exhibit distortion in time (or time elasticity) that requires using specific distance measures. In this work, we extend an algorithm proposed by Shokoohi et al. [35] by improving the performance of rule matching in the detection of energy consumption patterns. Nowadays companies are more and more equipped with sensors in order to trace losses of energy resources. Detecting dysfunctions from time series recorded by these sensors becomes a crucial problem for reducing energy consumption. Locating specific patterns related to dysfunctions in time series requires handling with time elasticity (i.e. distortion in time) of patterns. We propose a detection of predictive rules based on several variations of Dynamic Time Warping (DTW) and show the superiority of subsequence DTW [11]. We study now multivariate time series classification to predict dysfunctions as soon as possible.