How to train a time series dataset ?

I have a data set with 2 columns (the first column is related to currents of a battery and the second column indicates the time seconds correspond to the current). what I want to do is training the currents data according to the time (second columns).I convert my data set in form of time-series object:

(ts = timeseries(datavals,timevals)). And now I want to train this object. However I am not sure whether it is possible or not and how can I do it.

I would appreciate any further help.

Answer 1

[Majid Farzaneh]

The attached file may help you.

[Sara]

Dear Majid,

I appreciate your help.

[Majid Farzaneh]

You're welcome

Answer 2

[Majid Farzaneh]

Hi, what's your purpose for training this data? A neural network can predict something for you. For example you can train the network by current as input and time as output, then use the network, give a current to it and it provides a time for you. This is a single input/ single output network and you can use a feedforwardnet like this:

net=feedforwardnet(10); % 10 is number of neurons in the first layer
net=train(net,inputs,targets);

Now you have a trained network and you can use it like this:

   out=net(test_data)

Answer 3

[Sara]

Hello Majid, thanks for your answer, I just want to train current's behavior according to the corresponding time. So that in case of new data the system would recognize this data is current not for example voltage.

[Sara]

What I really want is to consider current and time as a unify input and my target is current

[Majid Farzaneh]

You're welcome. I've confused. Your trained network must recognize the input is current or not? or you want to estimate the value of current? If you want the first one, you have 2 possible outputs (1= is current, 0== is not current). But for this target you need more training data such as voltage or something else. So neural network can recognize patterns in currents vs times and something else vs times. Then when you give a new data to the network it can say your input is current or not. If you know Farsi, I can give you more help in Farsi.

[Sara]

Yes I can understand Persian. I would appreciate your help.

[Majid Farzaneh]

My comment in Persian considered as spam! Send your data if it's possible. I'll try to write a code for your purpose.

Answer 4

[Greg Heath]

Consider current as a function of time.

The rest should be straightforward using FITNET.

HOWEVER, YOU HAVE USED THE TERM TIMESERIES. THIS COULD IMPLY THAT EACH POINT BEYOND THE ITH DEPENDS ON THE PREVIOUS I VALUES AND NOT NECESSARILY EXPLICITLY ON TIME.

Thank you for formally accepting my answer

Greg

[Sara]

Thanks for you answer. could you please explain how should I consider it as a time function. the main problem I've faced is that considering current as time function. probably my problem is not time series, and I am in mistake.

Thank you in advance.