Take the data out after using Classification Learner App

3 visualizzazioni (ultimi 30 giorni)
Hi, I separated my data into training and testing data. I trained using the training data, then I tested using the testing data with the code below. It said "Unrecognized variable name 'Character'." And how can I take the data out after classification? Please help me. Thank guys.
testdata=readtable("ClassificationTestData.xlsx")
predictions = char(trainedModel.predictFcn(testdata))
% accuracy
iscorrect=predictions==cell2mat(string((testdata.Character)));
iscorrect=iscorrect(:,2);
accuracy = sum(iscorrect)*100/20;

Risposta accettata

Image Analyst
Image Analyst il 15 Dic 2021
Huy, here is a complete demo along with instructions in the comments
%==============================================================================
% TRAINING
trainingData=readtable("ClassificationData.xlsx")
% The first 4 columns are the inputs.
tPredictors = trainingData(:, 1:4);
% The last column is the "answer/ground truth".
tResponse = trainingData{:, end};
% Now run Classification Learner app. Start a new session from workspace, and
% Select tPredictors as the "Data set variable" and
% select tResponse as the response.
% I did "All quick to train" and found the Fine Tree model was the best.
% I highlighted it and clicked the Export button to export trainingModel to the workspace.
% Then save it to a .mat file from the command line like this:
% save('trainedModel.mat', 'trainedModel')
%==============================================================================
% TESTING
% Now run it on some test data once you have the trainedModel.mat file created.
% First read in the model from the .mat file
fileName = fullfile(pwd, 'trainedModel.mat');
if ~isfile(fileName)
message = sprintf('Model file not found:\n%s\n\nAre you sure you exported it from Classification Learner', fileName);
uiwait(errordlg(message));
return;
end
s = load(fileName)
trainedModel = s.trainedModel;
% Read in test data
testData=readtable("ClassificationTestData.xlsx")
testPredictors = testData(:, 1:4); % First 4 columns are the model inputs.
testResponse = testData{:, end}; % Last column is the ground truth (correct values).
% Predict the values.
numTestValues = size(predictions, 1)
% Compare predictions to ground truth.
itsCorrect = zeros(numTestValues, 1);
for k = 1 : numTestValues
if isequal(predictions{k}, testResponse{k})
itsCorrect(k) = 1;
end
end
% accuracy
accuracy = sum(itsCorrect)*100/numTestValues
  20 Commenti
Image Analyst
Image Analyst il 28 Dic 2021
Try this:
% TRAINING
trainingData=readtable("ClassificationData.xlsx")
% The first 4 columns are the inputs.
tPredictors = trainingData(:, 1:4);
% The last column is the "answer/ground truth".
tResponse = trainingData{:, end};
matFileName = fullfile(pwd, 'trainedModel.mat');
if ~isfile(matFileName)
warningMessage = sprintf('Model file not found:\n', matFileName);
uiwait(errordlg(warningMessage));
return;
end;
s = load(matFileName)
trainedModel = s.trainedModel
fn = fieldnames(trainedModel);
c = contains(fn, 'Classification');
regressionModelName = fn{find(c)}
% Read test data in to a table.
testData=readtable("ClassificationTestData.xlsx")
testPredictors = testData(:, 1:4); % First 4 columns are the model inputs.
testResponse = testData{:, end}; % Last column is the ground truth (correct values).
% Predict the values from the data in the tPredictors table.
predictions = trainedModel.predictFcn(testPredictors);
% Predict the values.
numTestValues = size(predictions, 1)
% Compare predictions to ground truth.
itsCorrect = zeros(numTestValues, 1);
for k = 1 : numTestValues
if isequal(predictions{k}, testResponse{k})
itsCorrect(k) = 1;
end
end
% accuracy
accuracy = sum(itsCorrect)*100/numTestValues
Everything is being classified as Cluster5. Is that what you expect?
Huy Cao
Huy Cao il 12 Gen 2022
Hi but acctually i have to use the test data from the classification learner app but i can import the testing data into the data set. can you help me thankss

Accedi per commentare.

Più risposte (3)

Cris LaPierre
Cris LaPierre il 15 Dic 2021
You appear to be trying to use a variable name (Character) that does not exist in your test data file. To me, it appears the available variable names are Power, WingSpeed, WindDirect, WindSpeed, and Cluster.
  2 Commenti
Huy Cao
Huy Cao il 15 Dic 2021
So you have any comment on that error please help me
Cris LaPierre
Cris LaPierre il 15 Dic 2021
I did. You need to use one of the actual variable names in your table, which appear to be Power, WingSpeed, WindDirect, WindSpeed, and Cluster.

Accedi per commentare.


Image Analyst
Image Analyst il 15 Dic 2021
You train using Power, WingSpeed, WindDirect, WindSpeed as the predictors, and Cluster as the response (ground truth).
Now when you go to use the trainedModel, you need to pass in only the predictors (columns 1-4 which are Power, WingSpeed, WindDirect, WindSpeed). Don't pass in Cluster as a predictor since the trained model won't be expecting that.
Your predictions are trainedModel.predictFcn(testdata). That's how you "can take the data out after classification". I'm not sure you need to cast it to char. And there is no column in your testdata table called "Character". Again, it needs to be Cluster since that's the column where the ground truth response for your test data is stored.
Attach trainedModel.mat if you need more help.
  3 Commenti
Huy Cao
Huy Cao il 15 Dic 2021
And in the WorkSpace I just got trainedModel 1x1 Struct
Image Analyst
Image Analyst il 15 Dic 2021
Save the model to a .mat file
save('trainedModel.mat', 'trainedModel');
then attach trainedModel.mat with the paper clip icon.
Character is not a field of the table so you need to extract the ground truth labels like this
testGroundTruth = testdata.Cluster;
iscorrect= predictions == testGroundTruth;
That may not work. I won't know unless I get the model. Or I'd have to train the model myself, but I don't know which model you chose.
Also make sure you didn't train your model with Cluster being one of the predictors as well as the response. Otherwise it will ignore your other 4 inputs since Cluster is, obviously, a perfect predictor of the response.

Accedi per commentare.


Huy Cao
Huy Cao il 15 Dic 2021
Hi thank you alot but when I added tPredictors into the data variables I couldn't add tResponse as a Response
  8 Commenti
Image Analyst
Image Analyst il 12 Gen 2022
I think for earlier versions it expects the response variable to be one of the columns in your predictors table. I was just working with a guy yesterday who had r2019b and we figured out that's what it was wanting.
Huy Cao
Huy Cao il 12 Gen 2022
@Image Analyst i mean I cant use the test data function in classification app. I use your code to import the data into workspace. Can you help me please
% TRAINING
trainingData=readtable("ClassificationData2.xlsx")
% The first 4 columns are the inputs.
tPredictors = trainingData(:, 1:4);
% The last column is the "answer/ground truth".
tResponse = trainingData{:, end};
testingData=readtable("ClassificationTestData2.xlsx")
tTesting=testingData(:,1:4);
ttestResponse=testingData{:,end};

Accedi per commentare.

Prodotti


Release

R2018a

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by