How can I run ten times with randomized data by using imageDatastore ? And how to do cross-validation by imageDatastore?

Hi, to test my project I split my data into data of train and data test... this is the code

[ImagesTrain,ImagesTest] = SplitData(ratio_App)

this is the code of function SplitData:

function [dataTrain,dataTest] = SplitData(ratio_App)

unzip(fullfile(matlabroot,'bin','ALLData.zip'));

data = imageDatastore('ALLData',...
    'IncludeSubfolders',true,...
    'LabelSource','foldernames');

[dataTrain,dataTest] = splitEachLabel(data,ratio_App);
dataTrain = shuffle(dataTrain);
end

I obtained the accuracy of my project 94% but when I want to validate it, I try to run my project 10 times by randomized data so I changed the function splitData like that:

function [dataTrain,dataTest] = SplitData(ratio_App)

    unzip(fullfile(matlabroot,'bin','ALLData.zip'));

    data = imageDatastore('ALLData',...
        'IncludeSubfolders',true,...
        'LabelSource','foldernames');

    [dataTrain,dataTest] = splitEachLabel(data,ratio_App,,'randomized');
    dataTrain = shuffle(dataTrain);
  end 

But by this function, I obtained accuracy in each run around 99% and 98% I don't understand why the results are not interpretable Could someone help me or show me where is the problem exactly??