I keep getting null results from NARX simulation.

Hi, I want to predict 16 timesteps using only the predicted inputs (not the targets). I got results when I use the function preparets with the targets preparets(net,testSeries,{},targetSeries);, but only getting null results when using it without it. i.e. preparets(net,testSeries,{});

The code follows:

inputSeries_5 = num2cell([acumulatedPrecipitation;temperature5cm]); targetSeries_5 = num2cell(soilMoisture5cm); testSeries_5=num2cell([[acumulatedPrecipitation,precipitationForecast];[temperature5cm,temperatureForecast5cm]]);

inputDelays = 1:16; feedbackDelays = 1:16; hiddenLayerSize = 10; net_5_Cm = narxnet(inputDelays,feedbackDelays,hiddenLayerSize); net_5_Cm.numInputs=3;

[inputs_5,inputStates_5,layerStates_5,targets_5] = preparets(net_5_Cm,inputSeries_5,{},targetSeries_5);

net_5_Cm.divideParam.trainRatio = 70/100; net_5_Cm.divideParam.valRatio = 15/100; net_5_Cm.divideParam.testRatio = 15/100;

net_5_Cm.trainParam.showWindow=0; view(net_5_cm);

[net_5_Cm,tr] = train(net_5_Cm,inputs_5,targets_5,inputStates_5,layerStates_5);

[xs,xis,ais,ts] = preparets(net_5_Cm,testSeries_5,{}); outputs_5 = net_5_Cm(xs,xis,ais);

Previsao5=[soilMoisture5cm(1:16),cell2mat(outputs_5)];

saida_5=outputs_5((end-16):end);

[Greg Heath]

VERY DIFFICULT TO FOLLOW YOUR NOTATION! Are you able to

1. Simplify your notation

2. Apply your code to a MATLAB dataset

 help nndatasets

 doc nndatasets

Thanks,

Greg