Supplementary material to the paper
Robust Monitoring of Time Series
with
Application to Fraud Detection
Peter Rousseeuw | Domenico Perrotta | Marco Riani | Mia Hubert | |
Department of Mathematics | Joint Research Centre, | Department of Economics and Management | Department of Mathematics | |
University of Leuven, | European Commission | University of Parma | University of Leuven, | |
Belgium | Italy | Italy | Belgium | |
peter@rousseeuw.net | domenico.perrotta@ec.europa.eu | mriani@unipr.it | mia.hubert@kuleuven.be |
Abstract
Time series often contain outliers and level shifts or structural changes. These
unexpected events are of the utmost importance in fraud detection, as they may
pinpoint suspicious transactions. The presence of such unusual events can easily
mislead conventional time series analysis and yield erroneous conclusions. In this
paper we provide a unified framework for detecting outliers and level shifts in short
time series that may have a seasonal pattern. The approach combines ideas from the
FastLTS algorithm for robust regression with alternating least squares. The double
wedge plot is proposed, a graphical display which indicates outliers and potential
level shifts. The methodology is illustrated on real and artificial time series.
Data used in the paper
Time series 1, regarding trade of “plants”: P_12119085.txt
Time series 2, regarding trade of “sugars”: P_17049075.txt
Simulated data: ysimout.txt
Matlab code used in the paper
routine LTSts.m. Link to the documentaion page of LTSts
routine simulateTS.m. Link to the documentaion page of simulateTS
routine forecastTS.m. Link to the documentaion page of forecastTS
routine wedgeplot.m. Link to the documenation page of wedgeplot
Remark: the above routines need the FSDA toolbox
Last modified 16/06/2018 08.01.10