convert knnclassify to fitcknn

please help me to convert knnclassify to fitcknn

x=readtable("datatraining.xlsx");
latih=x;
group=latih(:,3);
latih = [latih(:,1) latih(:,2)];
for i = 1 : 80
    y=readtable("datatesting.xlsx");
    sampel = y;
    test = [sampel(:,1) sampel(:,2)];
%sampel = [2.6136  0.1284 1.3017  -0.8089  0.0441  -0,2084];
    hasil=knnclassify(test,latih,group);
end
nama = "hasil KNN.xlsx";
hasil = [sampel(:,1) sampel(:,2) sampel(:,3) hasil];
xlswrite(nama,hasil);

[Walter Roberson]

Do not read files inside a loop -- not unless you are reading a different file each time.

The body of your loop does not use i, so you are doing the same thing each iteration.

[Frisda Sianipar]

Can you give the example sir in my code?

[Walter Roberson]

x=readtable("datatraining.xlsx");
y=readtable("datatesting.xlsx");
latih=x;
group=latih(:,3);
latih = [latih(:,1) latih(:,2)];
for i = 1 : 80
    sampel = y;
    test = [sampel(:,1) sampel(:,2)];
%sampel = [2.6136  0.1284 1.3017  -0.8089  0.0441  -0,2084];
    hasil=knnclassify(test,latih,group);
end
nama = "hasil KNN.xlsx";
hasil = [sampel(:,1) sampel(:,2) sampel(:,3) hasil];
xlswrite(nama,hasil);

except rewritten to use fitcknn as discussed in your other postings.

[Frisda Sianipar]

I've tried sir, but still get an error

[Walter Roberson]

Why did you go back to knnclassify() ? You were given conversion information at https://www.mathworks.com/matlabcentral/answers/505383-how-can-i-convert-knnclassify-to-fitcknn#comment_1500660

[Frisda Sianipar]

i have tried use fitcknn but still have error sir

[Walter Roberson]

And I already answered you at https://www.mathworks.com/matlabcentral/answers/505383-how-can-i-convert-knnclassify-to-fitcknn#comment_1502310

Answer 1

[Walter Roberson]

x = readtable("https://www.mathworks.com/matlabcentral/answers/uploaded_files/600525/datatraining.xlsx");
y = readtable("https://www.mathworks.com/matlabcentral/answers/uploaded_files/600520/datatesting.xlsx");

traindata = x{:,2};
traingroup = x{:,3};
testdata = y{:,2};
testgroup = y{:,3};

Mdl = fitcknn(traindata, traingroup,'Distance','euclidean','NumNeighbors',8,'Standardize',1,'BreakTies','nearest');
hasil = predict(Mdl, testdata);
nama = "hasil KNN.xlsx";
y.hasil = hasil;
writetable(y, nama)

does_it_match = strcmp(hasil, testgroup);
correct_percent = mean(does_it_match) * 100

correct_percent = 73.7500

[Frisda Sianipar]

it still got an error sir

[Frisda Sianipar]

x = readtable("https://www.mathworks.com/matlabcentral/answers/uploaded_files/600525/datatraining.xlsx");
y = readtable("https://www.mathworks.com/matlabcentral/answers/uploaded_files/600520/datatesting.xlsx");

will it be the same if I replace it with

x = readtable("datatraining.xlsx")
y = readtable("datatesting.xlsx")

[Walter Roberson]

x = readtable("datatraining.xlsx", 'readvariablenames', false);
y = readtable("datatesting.xlsx", 'readvariablenames', false);

[Frisda Sianipar]

Still have an error sir.

Error using classreg.learning.FullClassificationRegressionModel.prepareDataCR (line 192)
X must be a numeric matrix.

Error in classreg.learning.classif.FullClassificationModel.prepareData (line 487)
                classreg.learning.FullClassificationRegressionModel.prepareDataCR(...

Error in ClassificationKNN.prepareData (line 878)
                [e-mail](X,Y,varargin{:},'OrdinalIsCategorical',true);
                
Error in classreg.learning.FitTemplate/fit (line 213)
                    this.PrepareData(X,Y,this.BaseFitObjectArgs{:});

Error in ClassificationKNN.fit (line 863)
            this = fit(temp,X,Y);

Error in fitcknn (line 261)
    this = ClassificationKNN.fit(X,Y,RemainingArgs{:});

Error in KNNPleaseFix (line 7)
Mdl = fitcknn(traindata,
traingroup,'Distance','euclidean','NumNeighbors',8,'Standardize',1,'BreakTies','nearest');

[Walter Roberson]

filename = "datatraining.xlsx";
opt = detectImportOptions(filename);
x = readtable(filename, opt);

filename = "datatesting.xlsx";
opt = detectImportOptions(filename);
y = readtable(filename, opt);

traindata = x{:,2};
traingroup = x{:,3};
testdata = y{:,2};
testgroup = y{:,3};

Mdl = fitcknn(traindata, traingroup,'Distance','euclidean','NumNeighbors',8,'Standardize',1,'BreakTies','nearest');
hasil = predict(Mdl, testdata);
nama = "hasil KNN.xlsx";
y.hasil = hasil;
writetable(y, nama)

does_it_match = strcmp(hasil, testgroup);
correct_percent = mean(does_it_match) * 100

I tested this in you release to be sure it would work.

[Frisda Sianipar]

Thankyou so much sir, its works !!

Once again thankyou so much sir

[Walter Roberson]

There are several lessons here:

1. You had testing and training reversed
2. Predict all the data at one time instead of one at a time
3. You were working with all of the y data in every iteration of the loop, instead of extracting one y value
4. Do not re-read the data inside a loop if you can avoid doing so!
5. you were extracting the wrong data to use in the fitting
6. Use meaningful variable names
7. use writetable() instead of xlswrite()

[Frisda Sianipar]

Thankyou sir

[rashigha Shankar]

@Walter Roberson sir can you help me with a problem it is also same,the question link is given below

https://in.mathworks.com/matlabcentral/answers/1760240-can-somebody-please-help-me-how-to-convert-knncassify-function-to-fitknn

kindly check this and if you can please give a solution