Question

Q2) (Bonus) Use the MATLAB iris flowers data to classify the iris flowers only for the virginica and versicolor specimens. As you know, there are 4 measurement values for all the flowers. Which of the two measurement parameters give the best results for the classification? Attach your code and the figures and compare the parameter effects on the classification.

Answer 1

Iris plants chosen are

1. Iris Veriscolor

2. Iris Virginnica

Results reported with 1000 epochs with 4 neurons in the hidden layer and

10000 epochs with 4 neurons in the hidden layer.

KG = [input Data (26:30,:); inputData (106: 115, :), 4.% mup-dataset : Fisheries close alla clear alli dla ./ load dataset load fisherwis 4. Convert targets : : 1. veriscola", 23" Virginica" clase names Inique (species) ; for i= 1:2 class (stremp (species, classnames {i})) = 1; end class = class': % input data e Input Data (class meas]: / identification of classes of ing plant types as Coo); b = Colo]; ca Co o ili augTest 1 1. cross-validation : kfold (k = 10) manual i = 10; logize kx = 15x5 k1 - (input Data(155, :); inputDatal 56:65, :)) : k2 - Cinput data ( 6:10, :); inputData(06:75:-)) K3 = (input data (11515, :); input Data ( 76: 85, :)); k4 = Cinput Data (16:20, :); input data ( 86:95, :)]; Ks = [inputData (21:25,;); inputtata ( 96: 105, :));

K7 [input Data (31:35, :); input Data (116:125, k8 [input Data ( 36:40,:); inputate (126:135, :)] : K8 = [input Data (41:45, :); inputData (136:145,:)] ; K10 - (input Data (46 : 55,-) ; input data ( 146-150, :)); l. abole folds exchange fel se 1:3 if i==11. make train matrix (135x5) tranMatrix = (ke; k2;K3; k4; k5, ko'k7; 8;9]; test Matrix = k10 : end if i==2 of maketrain matrix (135x5) train Matrix (* 1; k2; k3;k 4; 5; KC; 17; K8 ; k 10]; testMatrix = kg; end -It do for i= 23 & 4, 5, 6, 7, 8, 9 if i= =10 .l. make train matrix (135x5) train Matrix = [k2; k3; k4; k5, K6; 17; 68, kq;k 10]; testmatrix k so end %% Start MLP 4. madrix for Trän (135x4) input Train train Matrix (:, 2:5); % target (135x1) target = tranMatrix (:,1); 1. nomalization of data input / input Train = mapmanmar (input Train); %-nomalization of target output Train 2010s ( size clasget, 1), 3);

// test network performance input Test - testMatrixt, 2:5); 1. (1584) target Test test Matrix(:,1): 3 for j= 1: size (target, 1) if target (j, 1) == 1 output TrainCj, ;) = a; end if target (3, 1) == 2 output Train (j, ;) = b; end if target ( in 1) ==3 output Train (j, ;) = C; end end %omatrix transposed input and output network input Train input Train ! ./. (4x135) output Train = output Train'; % (3x135) %.l. make network n-hidden-nodes = 4: net = feed forwardnet (n-hidden-nodes) net, trainParam, epochs = 2000 net, trainParam.max-fail = 500 net.brain Param.min - grad = 0.000000 0000 0000 1 net. divide Param, train Ratio = 1; net, divideParam, valRatio - 0 net, divideParam, test Ratio=0; net, layers $13. brausteifen a' tonsig': net layer (2} . transterfon 'Purelin'; // train network train (net, input Train, output Train); [net, xTest, yTest] 1. namalization of data-test (input Test): mapminmax % inputtet

output Test - zeros (Greeckooget Test; 1), 3); 781 3= 1; $ze ( torget Test, 1) if target Test ( ;, 1) = = 1 output Test (;, :) = a; end if target Test( J, 1) = -2 output Test (j :) = 6: end if target Test (j, 1) = 3 output Test (j, :) = C; end end % matrix transposed input & output network input Test input Test' . % (4X15) Output Test - cutput Test'; % (3815) lo network Prediction result result = SimCnet, input Test); (per con] = contusion (output Test, result): perTest - 100* (1-P); fprintf('fold=%d [ Accuracy 4.2.2 +1.1] In', i, potest; aug Test = aug Test & pertest; ol plot and Save confusion matrix x(i) = plot confusion (result, output Test); if i=1 saveas (x(i), confusion - 1. jpg"); ifraz if i=2 saveas (x(i), 'confusion - 2.jpg"); if 1-10 saveas (ali), confusion-10.jpg'); end end aug Test - aug Teselio, fprintf('network hat average (I fold =10): 134 -1.1. in augrest);

Results with 1000 spochs with 4 newons in the hidden lages Iris Plant Total No. of Test Samples No of classitted samples No of inclassified Samples 1. Inis versicolor 10 10 2. Inis Virginnica 10 O 10 with 10000 Epochs 4 newors in the hidden lages 1. Inis veriscolor lo 10 0 2 Iris Virginnice 10 0