How can i calculate forecast accuracy measures (MAD, MSE, and MAPE) using "Neural Network Time Series App" ?

i have a time series data set and i want to compute the accuracy measrues by using Neural Network Time Series App to make a comparison with other forecast model.

Answer 1

[Jaynik]

Hi Ahmed,

I understand that you want to calculate the accuracy measures like MAD (Mean Average Deviation), MSE (Mean Squared Error) and MAPE (Mean Average Percentage Error) using the “Neural Network Time Series App”.

Currently, however, you can only calculate MSE using the app. Once you load the network and data and train the model, the app automatically calculates the MSE and it can be seen inside the “Model Summary Pane”.

To calculate other measures using the obtained model, you can generate the code using “EXPORT > Generate Code”. Once you obtain the code, based on the values of variables ‘net’, ‘t’ and ‘y’, you can calculate each measures using the MATLAB inbuilt functions as follows:

% To convert cell array into matrix:
y = cell2mat(y);
t = cell2mat(t);

mean_squared = mse(net, t, y);
MAPE = mape(y, t);
MAD = mad(y);

You can read about the following functions here:

• mse: https://www.mathworks.com/help/deeplearning/ref/mse.html
• mad: https://www.mathworks.com/help/stats/mad.html
• mape: https://www.mathworks.com/help/matlab/ref/mape.html
• cell2mat: https://www.mathworks.com/help/matlab/ref/cell2mat.html

Hope this helps!