normalising and reverse normalising data

Question

NN 2020-12-8

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/685833-normalising-and-reverse-normalising-data

回答： Karan Nandankar 2020-12-28

For forecasting , I normalised the data before training for test and train data , then when i plot the rmse , i get rmse 0.2.but when i try to reverse the normalised values to match with orginal target of test data it shows high error.

i have used below code to normalise and denormalise :

[pn,ps] = mapminmax(XTrain); 
[tn,ts] = mapminmax(YTrain); 
[qn,qs] = mapminmax(XTest); 
[rn,rs] = mapminmax(YTest); 
net = trainNetwork(pn,tn,layers,options); 
YPred=predict(net,qn); 
YPred1 = mapminmax('reverse',YPred,ts);

what could be the reason ?

3 个评论
显示 1更早的评论隐藏 1更早的评论

NN 2020-12-8

i am using 2019b version.

i am doing forecasting Problem,

So i applied mapminmax function to rescale the data .

I referred below matlab documentation.

https://www.mathworks.com/help/deeplearning/ref/mapminmax.html

i am getting minimum rmse value with normalised output and test data .But when i reverse it to original scale , it shows different rmse..

Why is it happening .Kindly help..

i used below code:

[pn,ps] = mapminmax(XTrain);

[tn,ts] = mapminmax(YTrain);

net = trainNetwork(pn,tn,layers,options);

qn = mapminmax('apply', XTest,ps);

[rn,rs] = mapminmax(YTest);

YPred=predict(net,qn);

YPred1 = mapminmax('reverse',YPred,ts);

figure

plot(YPred1)

hold on

plot(YTest,'.-')

hold off

legend(["Observed" "Predicted"])

ylabel("SolarIrradiance")

title("Forecast with Updates")

rmse = sqrt(mean((YPred1-YTest).^2))

Star Strider 2020-12-8

Reference: My Comment to your previous Comment.

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Karan Nandankar 2020-12-28

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/685833-normalising-and-reverse-normalising-data#answer_585682

在 MATLAB Online 中打开

Hi,

Looks like you have used the wrong Process Settings in the variable 'YPred1'. As I can see you are using XTest as your independent variable for Model Prediction, and the corresponding dependent variable YTest is normalized with Process Setting parameter 'rs'. However, for the variable YPred1 you have mapped the predictions against 'ts' (which is for YTrain).

In order to reverse the normalization, you can change the Process Setting parameter in YPred1 from 'ts' to 'rs'.

YPred1 = mapminmax('reverse',YPred,rs);

You can refer to this documentation for more information.