Why do you need to predict this data for the first 75 seconds? can't you just remove it from your calculation?

Hint: Try to find out if the change is linear.

Best

Simple interpolation on the next n data items (e.g. 150 next data items) does not fit your requirement?

Note that generally, the initial state of the system is not subject to prediction, except the cases that the target of the research is to find those initial states; which I think is not your case.

Dear Ebrahim

You can use the curve fitting toolbox in matlab.

All you need to do is to introduce the inputs and output and select the proper method to do the fitting.

i think you can use time series prediction

Hi Ala AlAfeef

Thank you for your answer.

Unfortunately, I cannot remove data in the first 75 second as it plays an important role on my results. In addition, I am kind of sure that the relationship of data is not linear in that time span.

Cheers,

Dear Ali Rezaee

Thank you for your time.

Interpolation of data using its immediate neighbors came to my mind and I think it is reliable. It is also worth mentioning that finding out the behavior of my system in the initial stage (the first 75 seconds) is on of critical scopes of my experiment ...

Cheers

Dear Amir Hossein Salimi

Awesome. I tried using the time series method and relating the data in the time n to the data in the time n+1, n+2, or a combination of them. It works somehow. But, do you have any recommendation in this regard?

Kind regards,

Hello,

I am a bit rusty on this issue, but I think you could use least square to find a best fit for the data and use the missing x values to find their y's.

good luck

Have you checked splinefit function on Matlab?

It is a perfect piece-wise interpolation function, which respects continuity of the first and the second derivatives. It may suggest the best candidates for finding points out of the range.

hello,

please use any statistical forecasting techniques and check which one is suitable for your data.

good luck@

Dear Ebrahim

Hi

have you noticed that using your first 50 data, gives a really good fitted trendline in Excel, either power or logarithmic?

Hello Shahab Kabiri

Thank you for your answer.

Yes. I tried it before and the results were promising.

Due to its special behavior around zero, I rather not use logarithmic trend-line. However, the power function could fit my data (the first 5 minutes after missed data) appropriately. Then, given that there should be the same trend I could predict the missed data.

I thought it is a good idea to check the results with another method, like ANN and time series prediction. Unfortunately, ANN fails to predict the data out of the range and I could not make the time-series work on my data as well.

Thanks again for your time.

Cheers

dear Sir,

I hope to find you well, you could use any time series forecasting techniques, or use any curve fitting techniques to try and fit a function to your data.

All the best,

Sincerely yours,

saeid parsa