Input-Output Fitting problem with a Neural Network - Stock predication

Hi, I'm trying to predict stock values(in this case, the closing value). In the end I want to achieve the same result as this example, Example

I use Apples historical data over a year( historical data )

I will use the "close price" and its corresponding date, but I cannot figure out how to use "input" and "target" and what they mean.

    % This script assumes these variables are defined:
    x = simplefitInput; // 
    t = simplefitTargets;
    % Create a Fitting Network
    hiddenLayerSize = 10;
    net = fitnet(hiddenLayerSize,trainFcn);

    % setup Division of Data for Training, Validation, Testing
    net.divideParam.trainRatio = 70/100;
    net.divideParam.valRatio = 15/100;
    net.divideParam.testRatio = 15/100;
   % Train the Network
   [net,tr] = train(net,x);

   % Test the Network
   y = net(x);
   e = gsubtract(t,y);
   performance = perform(net,t,y)
   % View the Network
   %view(net)

If understand it correctly, x is the input of the system. But what is t ? So Im wondering: What kind of values do I give to x and t ?

Answer 1

[Greg Heath]

close all, clear all, clc, plt = 0

% NOTE: SOME SEMICOLONS REMOVED FOR DISPLAY PURPOSES

 % Read Useful Info
 %    help nndatasets
 %    doc nndatasets

 % x is the input
 % t is the training target for the output
 % y is the output
 % e is the error (e=t-y)

% Data Selection

 [ x ,t ] = simplefit_dataset; 
 whos
 [I N ] = size(x)           %[ 1 94 ]
 [O N ] = size(t)            %[ 1 94 ]

%Read more Useful Info

 %  help simplefit_dataset
 %  doc simplefit_dataset

 plt = plt+1, figure(plt)
 plot(x,t,'LineWidth',2)
 MSE00 = var(t',1)          % 8.3378 MSE Reference

% Using as many defaults as is reasonable:

net = fitnet;                 % net = net will reveal all defaults
[ net tr y e] = train(net,x); % tr = tr will reveal  training details 
%    y = net(x);
%    e = t - y
 MSE  = mse(e)      %  4.9880e-008  Mean-squared-error
 NMSE = MSE/MSE00   %  5.9824e-009  Normalized

% Coefficient of determination, Rsquared, is interpreted as the fraction of target variance that is modeled by the net. See Wikipedia for a decent discussion

   R2       = 1-NMSE           %  1.0000   Almost Perfect!

%See the training record,tr, for details of the default data division into training, validation and test subsets and training details

 tr = tr

Hope this helps.

Thank you for formally accepting my answer

Greg

Answer 2

[Greg Heath]

My previous answer is for regression and curve-fitting.

Unfortunately, I concentrated more on your given code than your written description.

You have a time-series problem and should be using the time-series function NARXNET; not the regression/fitting function FITNET.

 help narxnet
 doc narxnet

For MATLAB time-series example data

 help nndatasets
 doc nndatasets

I have posted tens of examples. Search the NEWSGROUP and ANSWERS using

 greg narxnet

and

 greg narxnet nncorr

For your case the output target is probably close or adjusted close. The inputs are the other tabulations. Try searching for

 narxnet stock
 narxnet forecasting

Hope this helps.

Thank you for formally accepting my answer

Greg