MASA: Motif-Aware State Assignment in Noisy Time Series Data

Abstract

Complex systems, such as airplanes, cars, or financial markets, produce multivariate time series data consisting of a large number of system measurements over a period of time. Such data can be interpreted as a sequence of states, where each state represents a prototype of system behavior. An important problem in this domain is to identify repeated sequences of states, known as motifs. Such motifs correspond to complex behaviors that capture common sequences of state transitions. For example, in automotive data, a motif of "making a turn" might manifest as a sequence of states: slowing down, turning the wheel, and then speeding back up. However, discovering these motifs is challenging, because the individual states and state assignments are unknown, have different durations, and need to be jointly learned from the noisy time series. Here we develop motif-aware state assignment (MASA), a method to discover common motifs in noisy time series data and leverage those motifs to more robustly assign states to measurements. We formulate the problem of motif discovery as a large optimization problem, which we solve using an expectation-maximization type approach. MASA performs well in the presence of noise in the input data and is scalable to very large datasets. Experiments on synthetic data show that MASA outperforms state-of-the-art baselines by up to 38.2%, and two case studies demonstrate how our approach discovers insightful motifs in the presence of noise in real-world time series data. CCS CONCEPTS• Computing methodologies → Cluster analysis; Motif discovery; Unsupervised learning.

Driver Behavior Detection in Time Series Decade review

“…The researchers in [95], in their trials to detect drivers' behavior time series patterns using their own dataset, during the driving lesson, RNN (GRU, LSTM) algorithms were used to detect patterns. The algorithm is unaffected by road factors, vehicle characteristics, location, or driver conduct.…”

“…The authors in [97] presented an approach for identifying maneuvers from vehicle telematics data, through motif detection in time-series. They used a modified version of the Extended Motif Discovery (EMD) method [95] that was applied to the UAH-DriveSet [98]. The first experiment attempted to detect acceleration and brakes from longitudinal acceleration time series, and the second attempted to recognize turns from lateral acceleration time series.…”

TripMD: Driving patterns investigation via Motif Analysis

“…Recently, Jain et al [2] proposed a general method to discover motifs in noisy time-series and, in one of their case-studies, they concluded that their method was capable of identifying turn maneuvers from automobile sensor data. This work further validates our claim that motifs extracted from driving sensor data are highly related to the actual maneuvers performed.…”

“…In this paper, we expand the work done in [1] by proposing TripMD, a complete motif extraction and exploration system that is tailored for the task of analyzing maneuvers and driving behaviors. Other authors have looked into the task of maneuver detection using timeseries motifs [3,2], however, none of these works propose a full system that extracts motifs from automobile sensor data and summarizes the information in a spaceefficient visualization.…”