ChangeDAR: Online Localized Change Detection for Sensor Data on a Graph

Given electrical sensors placed on the power grid, how can we automatically determine when electrical components (e.g. power lines) fail? Or, given traffic sensors which measure the speed of vehicles passing over them, how can we determine when traffic accidents occur? Both these problems involve detecting change points in a set of sensors on the nodes or edges of a graph. To this end, we propose ChangeDAR (Change Detection And Resolution), which detects changes in an online manner, and reports when and where the change occurred in the graph. Our contributions are: 1) Algorithm : we propose novel information-theoretic optimization objectives for scoring and detecting localized changes, and propose two algorithms, ChangeDAR-S and ChangeDAR-D respectively, to optimize them. 2) Theoretical Guarantees : we show that both methods provide constant-factor approximation guarantees (Theorems 5.2 and 6.2). 3) Effectiveness : in experiments, ChangeDAR detects traffic accidents and power line failures with 75% higher F-measure than comparable baselines. 4) Scalability : ChangeDAR is online and near-linear in the graph size and the number of time ticks.

[1]  Martin Hoefer,et al.  Online Independent Set Beyond the Worst-Case: Secretaries, Prophets, and Periods , 2013, ICALP.

[2]  David P. Williamson,et al.  A note on the prize collecting traveling salesman problem , 1993, Math. Program..

[3]  David P. Williamson,et al.  A general approximation technique for constrained forest problems , 1992, SODA '92.

[4]  Thomas M. Cover,et al.  Elements of Information Theory , 2005 .

[5]  Aristides Gionis,et al.  Event detection in activity networks , 2014, KDD.

[6]  Kenji Yamanishi,et al.  Detecting gradual changes from data stream using MDL-change statistics , 2016, 2016 IEEE International Conference on Big Data (Big Data).

[7]  Anil Vullikanti,et al.  Near-Optimal and Practical Algorithms for Graph Scan Statistics , 2017, SDM.

[8]  Nigel Collier,et al.  Change-Point Detection in Time-Series Data by Relative Density-Ratio Estimation , 2012, Neural Networks.

[9]  David S. Matteson,et al.  A Nonparametric Approach for Multiple Change Point Analysis of Multivariate Data , 2013, 1306.4933.

[10]  S. Massoud Amin.  U.S. grid gets less reliable [The Data] , 2011 .

[11]  Danai Koutra,et al.  DELTACON: A Principled Massive-Graph Similarity Function , 2013, SDM.

[12]  Miro Kraetzl,et al.  Using graph diameter for change detection in dynamic networks , 2006, Australas. J Comb..

[13]  Philip S. Yu,et al.  GraphScope: parameter-free mining of large time-evolving graphs , 2007, KDD '07.

[14]  A. Scott,et al.  A Cluster Analysis Method for Grouping Means in the Analysis of Variance , 1974 .

[15]  Xinwei Deng,et al.  Graph Scan Statistics With Uncertainty , 2018, AAAI.

[16]  Brandon Pincombea,et al.  Anomaly Detection in Time Series of Graphs using ARMA Processes , 2007 .

[17]  Alessandro Rinaldo,et al.  Changepoint Detection over Graphs with the Spectral Scan Statistic , 2012, AISTATS.

[18]  Zhengding Lu,et al.  Community mining on dynamic weighted directed graphs , 2009, CIKM-CNIKM.

[19]  Alessandro Rinaldo,et al.  Detecting Anomalous Activity on Networks With the Graph Fourier Scan Statistic , 2013, IEEE Transactions on Signal Processing.

[20]  Diane J. Cook,et al.  A survey of methods for time series change point detection , 2017, Knowledge and Information Systems.

[21]  Leto Peel,et al.  Detecting Change Points in the Large-Scale Structure of Evolving Networks , 2014, AAAI.

[22]  Hector Garcia-Molina,et al.  Web graph similarity for anomaly detection , 2010, Journal of Internet Services and Applications.

[23]  Ryan P. Adams,et al.  Bayesian Online Changepoint Detection , 2007, 0710.3742.

[24]  Piotr Fryzlewicz,et al.  Multiple‐change‐point detection for high dimensional time series via sparsified binary segmentation , 2015, 1611.08639.

[25]  Aristides Gionis,et al.  ACM SIGKDD Conference on Knowledge Discovery and Data Mining, KDD 2014, New York, NY, USA - August 24 - 27, 2014 , 2014 .

[26]  Jean-Philippe Vert,et al.  The group fused Lasso for multiple change-point detection , 2011, 1106.4199.

[27]  Hyun Ah Song,et al.  PowerCast: Mining and Forecasting Power Grid Sequences , 2017, ECML/PKDD.

[28]  Richard M. Karp,et al.  Reducibility Among Combinatorial Problems , 1972, 50 Years of Integer Programming.

[29]  Masashi Sugiyama,et al.  Change-Point Detection in Time-Series Data by Relative Density-Ratio Estimation , 2011 .

[30]  Daniel B. Neill,et al.  Non-parametric scan statistics for event detection and forecasting in heterogeneous social media graphs , 2014, KDD.

[31]  R D Zimmerman,et al.  MATPOWER: Steady-State Operations, Planning, and Analysis Tools for Power Systems Research and Education , 2011, IEEE Transactions on Power Systems.

[32]  Jeffrey D. Scargle,et al.  An algorithm for optimal partitioning of data on an interval , 2003, IEEE Signal Processing Letters.

[33]  Martin Pál,et al.  Algorithms for Secretary Problems on Graphs and Hypergraphs , 2008, ICALP.

[34]  Ambuj K. Singh,et al.  NetSpot: Spotting Significant Anomalous Regions on Dynamic Networks , 2013, SDM.

[35]  Piotr Indyk,et al.  A Nearly-Linear Time Framework for Graph-Structured Sparsity , 2015, ICML.

[36]  P. Fearnhead,et al.  Optimal detection of changepoints with a linear computational cost , 2011, 1101.1438.

[37]  Richard G. Lathrop,et al.  Urban change detection based on an artificial neural network , 2002 .

[38]  Ambuj K. Singh,et al.  Mining Evolving Network Processes , 2013, 2013 IEEE 13th International Conference on Data Mining.

[39]  Manuel Davy,et al.  An online kernel change detection algorithm , 2005, IEEE Transactions on Signal Processing.