# how to group time series data into stationary & non-stationary? [closed]

consider potential values as time series data points(4th column)

file<-read.table(filename.txt,skip=1) ts_data<-file$V4 #final time seris vetor please,give R script/functions to be used from any package,for follwing 1. to cluster these time seris into 2 groups--stationary & non-stationary(by programming & not by visually seeing ACF/PACF graphs) 2. how to find best model(AR/ARMA/ARIMA/SARIMA/GARCH,etc) to fit over these data for future forecasting ?,i don't want manually for each file,kindly suggest automated script 3. how to transform non-stationary models to get best fit model(i dont know how to get best P(autoregressive order),Q(moving average),D(differencing order)) - Could you, please, invest some time in formatting your question? Poorly worded questions are unlikely to attract attention (anywhere) and collect good responses (here), although @Emre generously tried to decipher the above and pointed to alternative (Matlab) solution. – chl May 16 '12 at 22:07 You are asking people to do an enormous amount of work for you. Please ask some specific question, but be prepared to do some work yourself. – Rob Hyndman May 16 '12 at 23:43 ## closed as not a real question by mbq♦Jun 4 '12 at 17:37 It's difficult to tell what is being asked here. This question is ambiguous, vague, incomplete, overly broad, or rhetorical and cannot be reasonably answered in its current form. For help clarifying this question so that it can be reopened, see the FAQ. ## 2 Answers I don't know if R has a package but the Machine Learning Group of TU Berlin has written a MATLAB toolbox for nonstationary time series decomposition called Stationary Subspace Analysis. Other references are • Satoshi Hara, Yoshinobu Kawahara, Takashi Washio, and Paul von Bunau. Stationary subspace analysis as a generalized eigenvalue problem. In Proceedings of the 17th International Conference on Neural Information Processing: Theory and Algorithms - Volume Part I, ICONIP’10, pages 422–429, Berlin, Heidelberg, 2010. Springer-Verlag. • Motoaki Kawanabe, Wojciech Samek, Paul von Bunau, and Frank Meinecke. An information geometrical view of stationary subspace analysis. In Artificial Neural Networks and Machine Learning ICANN 2011, volume 6792 of Lecture Notes in Computer Science, pages 397–404. Springer Berlin / Heidelberg, 2011. URL http://dx.doi.org/10.1007/978-3-642-21738-8_51 -  i am using "itsmr" package-autofit() function & grouped into 2 states,Am i correct? 1-method---> > t<-read.table(file.choose(),skip=1) > t<-t$V4 #final time series data vector > autofit(t) Error in arima(x, c(p, 0, q)) : non-stationary AR part from CSS another method----> j<-tryCatch(red_money<-autofit(ts(t)),error=function(e)e )  if(inherits(j,"simpleError")){print("Non-stationary")} but it takes more time – Sagar Nikam May 18 '12 at 14:56 Sorry, I don't use R. Furthermore, this is not the place to ask programming questions. – Emre May 22 '12 at 6:43

Yes, applying SSA is definitely an option.

This is the original reference:

http://dx.doi.org/10.1103/PhysRevLett.103.214101

And this paper explains how to get the most non-stationary directions:

http://dx.doi.org/10.1109/TNNLS.2012.2185811

-