Finally i wrote this code for wind speed prediction with 3 parameters, why does my code has different prediction for the same dataset on each run?

load('input.mat');
X = tonndata(inputData(:,(1:3)),false,false);

T = tonndata(inputData(:,4),false,false);

N = 100; % Multi-step ahead prediction

inputSeries  = X(1:end);
targetSeries = T(1:end);
inputSeriesVal  = X(end-N+1:end);
targetSeriesVal = T(end-N+1:end); 

delay = 1; %one hour
neuronsHiddenLayer = 10;

% Network Creation
net = narxnet(1:delay,1:delay,neuronsHiddenLayer);
[Xs,Xi,Ai,Ts] = preparets(net,inputSeries,{},targetSeries); 
net = train(net,Xs,Ts,Xi,Ai);
view(net)
Y = net(Xs,Xi,Ai); 
% Performance for the series-parallel implementation, only 
% one-step-ahead prediction
perf = perform(net,Ts,Y);
[Xs1,Xio,Aio] = preparets(net,inputSeries(1:end-delay),{},targetSeries(1:end-delay));
[Y1,Xfo,Afo] = net(Xs1,Xio,Aio);
[netc,Xic,Aic] = closeloop(net,Xfo,Afo);
[yPred,Xfc,Afc] = netc(inputSeriesVal,Xic,Aic);
multiStepPerformance = perform(net,yPred,targetSeriesVal);
view(netc)
figure;
plot([cell2mat(targetSeries),nan(1,N);
      nan(1,length(targetSeries)),cell2mat(yPred);]')
legend('Original Targets','Network Predictions')

Answer 1

[Jayanti]

Hi Morteza,

I understand that the predicted values are changing every time the code is executed.

The reason behind this can be random initialization of weights in the neural network. To fix this issue initialize the random number generator to a fixed seed before training the network. This will make sure that initial weight assignment remains same each time you run the code.

You can fix this by using

rng(seed_value)

You can specify any seed value you want.

I also reproduced this scenario using custom dataset and by fixing the seed value, the code resulted in the same prediction values for each run.

I am also attaching official MathWorks documentation on “rng” for your reference:

https://www.mathworks.com/help/matlab/ref/rng.html

Hope it helps!