slipting sata manually into train, validation and test sets, then using fitcnet and testing accuracy

Question

AIKATERINI NINOU 2022-8-21

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1782755-slipting-sata-manually-into-train-validation-and-test-sets-then-using-fitcnet-and-testing-accuracy

回答： KSSV 2022-8-21

Hey guys!

So, I am trying to design a neural network to classify subjects into three classes via fitcnet. There is a txt file that consists of 1092 rows and 14 collums (1092 x 14). The 14th collumn has the class of the sets either A, B or C and I specificly want the first 450 x 14 of them to be the training set , the next 150 x 14 to be the validation data and the rest of them to be the test set. Here is my code:

clc; clear all;

creditrating = readtable("STATSTOTAL.txt");

creditrating = removevars(creditrating,"Row");

creditrating.class = categorical(creditrating.class, ...

["A","B","C"],"Ordinal",true);

rng("default") % For reproducibility of the partition

creditTrain = creditrating(1:450 ,:);

creditVal=creditrating (451:600, :);

creditTest= creditrating (601:end, :);

It works fine!

****The thing is, how do I use the fitcnet function and how do I test the accuracy now that I have the three sets?

Without the validation data I did this following the example in matlab-help: