Nonlinear Time Series Analysis Tools
This is a Web Page where one can find Fortran Codes to compute some important quantifiers based on chaos theory and used for Nonlinear Time Series Analysis. These codes have grown out of the collaborative research of our research group consisting of Myself, R. Misra (IUCAA, Pune) and G. Ambika (IISER, Pune) over the last few years. We acknowledge the financial support in the form of a Research Grant received from The Department of Science and Technology (DST), Govt. of India (Grant NO. SR/S2/HEP-11/2008) and Science & Engineering Research Board (SERB), Govt. of India (Grant NO. SR/S2/HEP-27/2012) for this research.
Important: All the Codes presented here are Copyright Protected and can be used only for Academic purposes after getting prior permission from the authors.
The Copyright is jointly shared by K. P. Harikrishnan, R. Misra and G. Ambika.
The quantifiers for which the codes have been developed are Correlation Dimension (D2), Correlation Entropy (K2), Generalised Dimensions (D-q) and the f-alpha spectrum. For computing D2 and K2, codes have been written based on both GP algorithm and Box Counting algorithm. The advantage of the latter is that they are much faster and hence can be employed to compute D2 and K2 from time series involving much larger number of data points. In particular, they are useful for hyperchaotic and high dimensional systems.
An important aspect of all the codes presented here is that they are automated. It means that once the input time series is given, the code computes the relevant quantifier without requiring any intermediate subjective analysis. Along with every code in the package, a "read me" file and sample "input" and "output" files have also been provided for verification.A separate folder with some Synthetic and Real World data used for our analysis have also been included.
My Research student Rinku Jacob is also involved in developing the Codes for the construction and analysis of Recurrence Networks from Time Series.
Kindly Note: We do not claim that these codes are absolutely fool-proof and optimized for all the different types of time series data from the real world with different data lengths and involving various types of noise. Hence, while applying these codes, every one is requested to first confirm that the results obtained are consistent with other methods and analysis before publishing their work.
We are sure that there is scope for further improvement. Any comments, suggestions for improvement are welcome.