How to improve the performance of my neural network

I am using matlab nprtool to train a neural network for classification. I have 367 data samples of 32 inputs and 2 classes. 207 belongs to class A and the remaining 160 belongs to the other class.I have changed the number of hidden layer many times but the one that produced a manageable result is 15. However the result is not satisfactory enough. 55 test sample was given to the trained net, 34 were classified correctly and 21 wrongly. I have a normalized dataset but i did not use it because i saw online that nprtool does normalization to all dataset by default. Please,how can i increase the performance of the network. Also, i changed training function from trainscg to trainlm but that has not helped. Below is the code generated by matlab.

    %Input367_Decimal - input data.
    target367 - target data.

    inputs = Input367_Decimal;
    targets = target367;

    % Create a Pattern Recognition Network
    hiddenLayerSize = 15;
    net = patternnet(hiddenLayerSize);

   % Choose Input and Output Pre/Post-Processing Functions
   % For a list of all processing functions type: help nnprocess
   net.inputs{1}.processFcns = {'removeconstantrows','mapminmax'};
   net.outputs{2}.processFcns = {'removeconstantrows','mapminmax'};
   % Setup Division of Data for Training, Validation, Testing
   % For a list of all data division functions type: help nndivide
   net.divideFcn = 'dividerand';  % Divide data randomly
   net.divideMode = 'sample';  % Divide up every sample
   net.divideParam.trainRatio = 70/100;
   net.divideParam.valRatio = 15/100;
   net.divideParam.testRatio = 15/100;

   % For help on training function 'trainlm' type: help trainlm
   % For a list of all training functions type: help nntrain
   net.trainFcn = 'trainlm';  % Levenberg-Marquardt

   % Choose a Performance Function
   % For a list of all performance functions type: help nnperformance
   net.performFcn = 'mse';  % Mean squared error

   % Choose Plot Functions
   % For a list of all plot functions type: help nnplot
   net.plotFcns = {'plotperform','plottrainstate','ploterrhist', ...
  'plotregression', 'plotfit'};

   % Train the Network
[net,tr] = train(net,inputs,targets);

   % Test the Network
   outputs = net(inputs);
   errors = gsubtract(targets,outputs);
   performance = perform(net,targets,outputs)

    % Recalculate Training, Validation and Test Performance
    trainTargets = targets .* tr.trainMask{1};
   valTargets = targets  .* tr.valMask{1};
   testTargets = targets  .* tr.testMask{1};
   trainPerformance = perform(net,trainTargets,outputs)
   valPerformance = perform(net,valTargets,outputs)
testPerformance = perform(net,testTargets,outputs)

    % View the Network
    view(net)

    % Plots
    % Uncomment these lines to enable various plots.
    %figure, plotperform(tr)
    %figure, plottrainstate(tr)
    figure, plotconfusion(targets,outputs)
    %figure, ploterrhist(errors)

Any help will be much appreciated. Thanks

[mizuki]

In this case, the sample size might have correlations each other and be small to classify correctly.

Did you check with other classification methods such as SVM and decision tree? Classification Learner App would help you.

[Omotayo Asiru]

Thanks for your response. I have not checked with other classification methods but i will read about them now. What do you mean by "sample size might have correlations each other". Please explain. Thanks

[mizuki]

I meant the sample size might be small, which depends on the dataset. This happens especially when some samples are similar.

[Omotayo Asiru]

Ok, i get your point now. I read about the classification learner app online but could not get any website to download it. Do u happen to know how I can download it ? maybe a link to its download. Thanks

[Greg Heath]

Your initial post says number of number of hidden layers where you should have said number of hidden layer NODES.

HTH

Greg

Answer 1

[Greg Heath]

The code you posted is too complicated for your first time out. See the examples in the documentation using the commands

 help patternnet
 doc patternnet

Then try the example data in

 help nndatasets
  doc nndatasets

Finally, you can search BOTH the NEWSGROUP and ANSWERS using

 greg patternnet.

Hope this helps.

Thank you for formally accepting my answer

Greg

[Omotayo Asiru]

Thanks for your response. I tried the examples(iris dataset for flower classification problem ,wine and cancer classification also) before i proceeded to d dataset for my work. However,those examples produced very good output,I got some that are almost 100% accurate. That is why i need a way to improve the performance of my network too. Maybe there is any default settings that i could change. Thanks