Feature Flocks : accurate pattern discovery in multivariate signals

Abstract

Thanks to the rise of wearable and connected devices, sensor-generated time series comprise a large and growing fraction of the world's data. Unfortunately, extracting value from this data can be challenging, since sensors can only report low-level signals (e.g., acceleration), not the high-level phenomena that are typically of interest (e.g., gestures). We introduce a technique to bridge this gap by automatically learning to identify real-world events in low-level data with no human labeling. By identifying "flocks" of features that repeat in the same temporal arrangement, we learn to recognize such diverse phenomena as human actions, power consumption patterns, and spoken words with up to 96% precision and recall. Our method is fast enough to run in real time and assumes only minimal knowledge of which variables are relevant or how long patterns are. Our evalation uses numerous publicly available datasets and over 1 million samples of sensor data in which we manually labeled ground truth.