Re: HMM applied

Mario wrote:
I'm working on some time series. My need is to substitute the observations with lines of trend doing it computationally, because I have to work on a lot of time series.
I read something about the use of Hidden Markov Model to reach this objective ("Tutorial PM2 - Time series similarity measures" of Prof Gunopulus and Das, page 52-55).

I'm looking for a document that describes in detail the algorithm to use to translate it in a program.

Could someone help me?

Thanks in advance.

Mario Sacco
What I'm looking for is an algorithm for segmenting time series. Could be this is not the right ng. I'll be glad if someone would give me a pointer to the right group too.
Mario
.

Relevant Pages

  • Re: HMM applied
    ... observations with lines of trend doing it computationally, ... Mario Sacco ... What I'm looking for is an algorithm for segmenting time series. ...
    (sci.math.num-analysis)
  • Discrete-time finite-state Markov model
    ... My need is to substitute observations with multiple lines of trend (of variable length, related to the time series values). ... Now I'm looking for a document that describes in detail the algorithm to be used, so I could translate it in a program. ...
    (comp.dsp)
  • Re: wrong R-Squared value??
    ... get 200 data points (x is a time series; ... What do you mean by a simple trend? ... predicting from the original 200 points. ... original equation. ...
    (sci.stat.math)
  • Re: test for outlier in time series
    ... taking into consideration the variance and the trend ... Or should I go and look somewhere else, e.g. in time series modelling? ... "outliers" in cross-sectional data are Pulses or one time events. ... Intervention Variables possible. ...
    (sci.stat.math)
  • Re: test for outlier in time series
    ... my data is a time series with a linear trend and some noise, ... I thought about doing a simple linear regression and somehow using the ... Is there a standard measure for testing outliers in such a linear ... Or should I go and look somewhere else, e.g. in time series modelling? ...
    (sci.stat.math)