how to fix NaN value?

5 次查看（过去 30 天）

Suaad Al-Hussainan 2022-6-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1736100-how-to-fix-nan-value

评论： Suaad Al-Hussainan 2022-6-8

% Import Data
data = readmatrix('matrix 1.csv');
x = data(:,1:3);
y = data(:,4);
m = length(y);
% Visualization of data
histogram(x(:,3),10);
plot(x(:,3),y,'o');
% Normilize the features and transform the output
y2 = log(1+y);
for i = 1:3
    x2(:,i) = (x(:,i)-min(x(:,i)))/(max(x(:,i))-min(x(:,i)));
end
histogram(x2(:,1),10);
% Train the Artificial Neural Network (ANN)
xt = x2';
yt = y2';
HiddenLayerSize = 10;
net = fitnet(HiddenLayerSize);
net.divideParam.trainRatio = 90/100;
net.divideParam.valRatio = 50/100;
net.divideParam.testRatio = 0/100;
[net,tr] = train(net, xt, yt);
% Preformance of the ANN network
yTrain = exp(net(xt(:,tr.trainInd)))-1; % GIVES NaN value
yTrainTrue = exp(yt(tr.trainInd))-1;
MSE = mean((yTrain - yTrainTrue).^2); %MSE gives NaN value
RMSE = sqrt(mean((yTrain - yTrainTrue).^2)); %RMSE gives NaN value
RealPercentageOfDestruction = exp(net(xt(:,tr.trainInd)))-1;%Real Percentage Of Destruction gives NaN value
yVal = exp(net(xt(:,tr.valInd)))-1; GIVES NaN value
yValTrue = exp(yt(tr.valInd))-1;
MSE1 = mean((yVal - yValTrue).^2); %MSE gives NaN value
RMSE1 = sqrt(mean((yVal - yValTrue).^2)); %RMSE gives NaN value

6 个评论
显示 4更早的评论隐藏 4更早的评论

Jan 2022-6-8

在 MATLAB Online 中打开

The question is vague. Wheher do you observe NaN values? What does "fixing" mean?

data = [30, 9.6, 0,      61.7; ...
        40, 9.6, 53.885, 58.6; ...
        50, 9.6, 61.725, 55.7; ...
        60, 9.6, 62.555, 60.4; ...
        70, 9.6, 63.415, 54.4];
x = data(:,1:3);
for i = 1:3
    x2(:,i) = (x(:,i)-min(x(:,i)))/(max(x(:,i))-min(x(:,i)));
end

The first NaNs appear here: for i=2, the max and min values are the same, so you divide by zero. How do you want to "fix" this? It is the correct result in a mathematical sense.

Suaad Al-Hussainan 2022-6-8