why am i getting unsatisfied result with same Training and Testing data in Narx time series network?

I am planning to predict the temperature as a output while providing the input parameter as a current. I have uploaded my time series data data. i am training with same data set and testing on same data. but i am getting a unsatisfied result. Beforehand i have trained with this time series data set and test with another time series data set, but i did not get the result, which i was expecting. Now i am trying to Traing and Testing with same data set to find out if my Model is learning with this dataset or not, still it is not working. i am not getting satisfied result, even i am trying to train and test same data set. why? here is my Narxnet code

Data = readtable('Data.xlsx');

Current = table2array(Data(:,"I_DC")); % Input Value (Current)

Output = table2array(Data(:,"T_AGG")); % Target value (Temperature)

X_Train = tonndata(Current(1:end,:),false,false);

Y_Train = tonndata(Output(1:end),false,false);

trainFcn = 'trainlm'; %

inputDelays = 1:2;

feedbackDelays = 1:2;

hiddenLayerSize = 5;

net = narxnet(inputDelays,feedbackDelays,hiddenLayerSize,'open',trainFcn);

[x,xi,ai,t] = preparets(net,X_Train,{},Y_Train);

[net,tr] = train(net,x,t,xi,ai);

% Test the Network

[y,Xf,Af] = net(x,xi,ai);

e = gsubtract(t,y);

performance = perform(net,t,y)

% closed loop the same data for Testing multiple time step prediction

[netc,Xic,Aic] = closeloop(net,xi,ai);

view(netc)

Yc = netc(X_Test,Xic,Aic);

figure, plotregression(Y_Test,Yc)

figure, plotresponse(Y_Test,Yc)

It would be very helpful for me, if you could help me with this issue? I am not getting what i am doing wrong. or data is not appropriate to learn for Narxnet time series? Any suggestion would be helpful....Thanks in advance

Answer 1

[Ganesh]

Hi @Hemant,

Let's try and see the correlation between Current and Temperature.

Data = readtable('Data.xlsx');
Current = table2array(Data(:,"I_DC")); % Input Value (Current)
Output = table2array(Data(:,"T_AGG")); %  Target value (Temperature)
corrcoef(Current,Output)

ans = 2x2

1.0000 -0.0561 -0.0561 1.0000

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

Correlation of less -0.056 is pretty low, the Target variables show very low dependency with the Input Variables i.e. the externally determined variable shows low influence.