getting error about wrong input argument

Question

Haleema Ahsan 2020-7-2

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/558436-getting-error-about-wrong-input-argument

评论： Walter Roberson 2021-10-21

i have tried following code for upsampling of data with input values

smote('C:\Users\Haleema\Desktop\skin matlab\orignal', 8,{327,514,1099,115,1113,6705,142})

but got error

Error using bar (line 127)

Input arguments must be numeric, datetime, duration or categorical.

Error in barh (line 44)

h = bar(varargin{:});

Error in smote (line 22)

barh(sortedIDX);

function allData_smote = smote(allData, k,sortedIDX)
% mySMOTE  Synthetic Minority Oversampling Technique. A technique to
% generate synthetic samples as given in: https://www.jair.org/media/953/live-953-2037-jair.pdf
%   Usage:
%   X_smote = mySMOTE(X, N, k) 
%   
%   Inputs:
%   allData: Original dataset
%   k: number of nearest neighbors to consider while performing
%   augmentation
%   sortedIDX: sorted labels
% 
%   Outputs:
%   X_smote: augmented dataset containing original data as well.
%   
%   See also datasample, randsample
%% plot the bar plot for number of classes
figure
barh(sortedIDX);
ylabel('number of classes-->')
xlabel('Sampels in each class-->')
title('Original imbalance data distirbution')
%% number of each classes
labels=allData(:,end);
class=unique(sortedIDX);
for ii=1:numel(class)
    classNo(ii)=numel(find(labels==class(ii)));
end
%% required addon samples in each minority class
%add on samples will be calculated by taking the difference of each
%classSamples with highest number of class samples
[maximumSamples,sampleClass]=max(classNo); % number of maximum samples
for ii=1:numel(class)
    samplediff(ii)=maximumSamples-classNo(ii);
    N (ii) = ceil(samplediff(ii)/ 100);
    
end
%% oversample the minority classes
allData_smote=[];
for ii=1:numel(class)
    X=allData(labels==class(ii),:);
    T = size(X, 1);
    X_smote = X;
    for i = 1:T
        y = X(i,:);
        % find k-nearest samples
        [idx, ~] = knnsearch(X,y,'k',k);
        % retain only N out of k nearest samples
        idx = datasample(idx, N(ii));
        x_nearest = X(idx,:);
        x_syn = bsxfun(@plus, bsxfun(@times, bsxfun(@minus,x_nearest,y), rand(N(ii),1)), y);
        X_smote = cat(1, X_smote, x_syn);
    end
    allData_smote=cat(1,allData_smote,X_smote);
end
%%
balanced_sortedIDX=allData_smote(:,end);
figure
barh(balanced_sortedIDX);
ylabel('number of classes-->')
xlabel('Sampels in each class-->')
title('Balanced data distirbution')
%% randomize the data
shuffleindex=randperm(size(allData_smote,1));
allData_smote=allData_smote(shuffleindex,:);
end

5 个评论
显示 3更早的评论隐藏 3更早的评论

Geoff Hayes 2020-7-3

Haheema - how much of the above code have you modified from the original mySmote.m? If you are not the author of the above smote function, then you may want to contact whomever wrote it so that you can understand what the correct inputs should be.

Haleema Ahsan 2020-7-3

ok thanks

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2020-7-3

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/558436-getting-error-about-wrong-input-argument#answer_460334

在 MATLAB Online 中打开

smote('C:\Users\Haleema\Desktop\skin matlab\orignal', 8, [327,514,1099,115,1113,6705,142])

Not a cell array of numeric values, a numeric array of values.

6 个评论
显示 4更早的评论隐藏 4更早的评论

Walter Roberson 2020-7-3

The code appears to be adapted from https://www.mathworks.com/matlabcentral/fileexchange/70315-smote-over-sampling

The first parameter to smote() must be a 2D numeric array. The last column must have class number information.

The second parameter must be a numeric scalar indicating the number of nearest neighbours to consider.

The third parameter must be a numeric column vector that lists all of the possible classes once.

This is not exactly what the original author had in mind, but it gives the least bad graph without interfering with the computation. The bar graph that is produced will not be useful. The third parameter can repeat entries without it interfering with the computation, but the graph would become more useless. (In order for the graph to be useful, the third parameter would have to be a 2D array, and each class label would have just happen to exactly match the number of rows that had that same label, and the number of columns would have to match the number of different classes.) To be honest, I think the processing of the third parameter is broken, and I cannot figure out what it was intended to be.

smote() does not usefully accept a character vector as the first parameter -- it would treat it as being a 1 x something numeric array.