(Spatio-temporal Cluster-based Vector AutoRegressive model)

cVAR





 
A short description 
The distribution package 
Related publications 
Authors 
Contact 



A short description

Forecasting in geophysical time series is a challenging problem with numerous applications. The presence of correlation (i.e. spatial correlation across several sites and time correlation within each site) poses difficulties, with respect to traditional modeling, computation and statistical theory. This paper presents a cluster-centric forecasting methodology that allows us to yield a characterization of correlation in geophysical time series through a spatio-temporal clustering step. We present a novel spatio-temporal forecasting methodology, called cVAR, which input a dataset and consists of a pipeline of three algorithmic steps: (1) Computation of a clustering pattern, partitioning the geo-sensor system into distinct clusters on the ground of the spatial and temporal similarity of the time series. A peculiarity of our approach is the definition of a novel spatio-temporal dissimilarity, which includes both time series dissimilarity and spatial distance as separate contributions. (2) A feature expansion mechanism defining a system that couples an observation site with a multivariate system including the target time series observed at the site and a number of cluster-defined spatially-coupled time series. (3) Application of a stationary multivariate VAR time series model, in order to analyze the structure of this system of variables and construct an accurate predictor of the target time series with the help of the spatially coupled variables.

The distribution package

The cVAR algorithmic pipeline is implemented in a R environment (distributed through the CRAN network). It also uses the Java implementation of the local choice of the smooting parameter α, according to the AFFECT framework introduced in Xu et al. (2014).

rar Description
cVAR This rar bundle contains: (1) preProcess.R (and auxiliary files), that imports and normalizes training time series, computes spatio-temporal dissimilarities, performs clustering and calculates five-additional time series accounting for the spatial and temporal aware information enclosed in the clustering pattern; (2) cVar(tceqExample).R, that imports and normalizes test data using parameters determined in the training data, provides a function for estimating the seasonal period, performs VAR model order choice, parameter estimation and forecasting, and calculates the forecast accuracy (script to be used with the TCEQ dataset); (3) cVar(mesaExample).R, script to be used with the MESA dataset. For more information, please refer to the README file included within the bundle.
Local Alpha Calculator This rar bundle contains: (1) localAlpha.jar (.java + .class) that allows us to determine local α and auxiliary files, estimated according to the theory of Xu et al. (2014); (2) dataset directories that collect scaled data, as well as dissimilarity matrix and local alpha computed by using localAlpha.jar; (3) example of .bat file to run localAlpha.jar.

Warning: The cVAR software is free for evaluation, research and teaching purposes, but not for commercial purposes. The software is provided "as is" without warranty of any kind, either expressed or implied.
 
Please cite as: S. Pravilovic, M. Bilancia, A. Appice, D. Malerba. Using multiple time series analysis for geosensor data forecasting Information Sciences 0 0 0 (2016) 122
 

Top of this page


Related publications

S. Pravilovic, A. Appice, D. Malerba. An Intelligent Technique for Forecasting Spatially Correlated Time Series (2013) in: M. Baldoni, C. Baroglio, G. Boella, R. Micalizio (eds.), AI*IA 2013: Advances in Artificial Intelligence SE - 39, vol. 8249 of Lecture Notes in Computer Science, Springer International Publishing, pp. 457-468.
 
S. Pravilovic, A. Appice, D. Malerba. Integrating cluster analysis to the ARIMA model for forecasting geosensor data (2014) in: T. Andreasen, H. Christiansen, J.-C. Cubero, Z. Ra (eds.), Foundations of Intelligent Systems, vol. 8502 of Lecture Notes in Computer Science, Springer International Publishing, pp. 234-243.
 
K. S. Xu, M. Kliger, A. O. Hero III. Adaptive evolutionary clustering (2014) Data Mining and Knowledge Discovery, 28 (2), pp. 304-336.
 

Top of this page


Authors

  • Annalisa APPICE
  • Massimo BILANCIA
  • Sonja PRAVILOVIC
  • Donato MALERBA


    Contact

    Massimo Bilancia massimo.bilancia@uniba.it
    Annalisa Appice annalisa.appice@uniba.it
     

    Top of this page