I would like to use time-series to predict the behavior of a system with stochastic behavior. Since I am not quite familiar with this topic, could anyone point me to some good tutorial for this purpose? The nature of my inquiry is very practical, and thus I only need a minimal amount of theoretical background in order to understand it and a fairly good number of examples.
- Anybody can ask a question
- Anybody can answer
- The best answers are voted up and rise to the top
You might want to look into fitting a ARMA model to a data sample and then use this model to predict the system. Though I haven't been able to find a good and practical tutorial.
An alternative way, instead of using ARMA model, is to use the neural networks. Here is a tutorial
You can put your time series data X(t-1), X(t-2), X(t-3)... as input and the network will predict X(t).
For the practical part, it is very very simple to implement and train a neural network in matlab, just collect the data and use the nftool tool.
The book by Box, Jenkins and Reinsel is an excellent text to show you how to fit ARIMA and seasonal ARIMA time series models. It is very practical and not too theoretical. The little monograph by Chatfield is more concise and also good. Dave Reilly's Autobox software is a tool that can generate such models for you and it will even handle interventions. Dave Reilly is a member of this group "IrishStat". He can tell you about it in more detail.