How can I get my feature name from the location of matrix?

I have 3 input manifest files, each has 137 different features name to examine which save in 'Dataset1_Permission.txt'. If it exists in the files, it is 1, if not, will be 0. I will use the 137X3 logical vector to perform some calculation as I need to find the top 10 feature names in both 3 files. Now, I have sum each feature row by row and sort them to find the top 10 maximum features that exist in both 3 files. I can find the top 10 and retrieve the location of matrix but how can I get the features name of it? I can't figure out how to do so. Below is my sample code:

 function [features1]  = permission(~)

    for app=1:3

    text1 = fileread(sprintf('C:/Users/ASUS/Desktop/FYP/B%d/AndroidManifest.xml',app)); % read input manifest file
    text2 = regexp(fileread('Dataset1_Permission.txt'), '\r?\n', 'split'); % read android permission list dataset
    saveValue1 = [];

    matchStr = regexp(text1,'\"android.permission.\w*\"','match'); %extract the keyword in manifest file
    saveValue1 = [saveValue1 matchStr];
    no_duplicates1 = unique(saveValue1);
    str1 = strjoin(no_duplicates1);
    features1{app} = ~cellfun(@isempty,regexp(str1,text2));
    disp(no_duplicates1);

    end

    for app=1:length(features1)

    features1{app}=features1{app}';

    end

features1 = cell2mat(features1);
total =sum(features1,2);
[max_sorted,max_located] = sort( total, 'descend' );
n = max_sorted(1:10);
p = max_located(1:10);
disp(n)
disp(p)

 end

Answer 1

[Walter Roberson]

text2(p)

By the way, I would appreciate it if you stopped doing those strcat and switched to having a variable holding the directory name, and using fullfile(). Repeating the location construction like you are is error prone. It is also not at all portable. If you were to give the code to anyone else (like us volunteers) they would first have to spend their time searching the code and editing the location information repeatedly to match their own directory names. For example I cannot just change the username ASUS in multiple places because I do not run on MS Windows.

I already showed you the code to fix this issue... More than once.

[ai ping Ng]

Really appreciate your help, and so sorry for the inconvenience I have made. It is my fault that I didn't notice mine issue. This is my first year, first semester, I m still a newbie for coding. Thanks for advising me.

Answer 2

[ai ping Ng]

After done for feature selection coding, I get my top 10 features also, but instead of getting integer numbers like before 1 4 7 90..., it gave me floating numbers 0.3033 0.2899 0.1277 0.0011..... So now, when I used text2(p), it came out the error as 'Subscript indices must either be real positive integers or logicals.' What can I do now to get back my feature name of it?

[Walter Roberson]

How are you doing the feature selection coding?

[ai ping Ng]

% Sort features according to pairwise correlations

corrMatrix = corr(features1,'rows','pairwise');

% Get the scores according to minimum correlations

scores = min(corrMatrix,[],2);

% Exclude NaN and sort accordingly

sorted_score = sort( scores(~isnan(scores)) );

p = text2(sorted_score );

[Walter Roberson]

[sorted_score, sort_idx] = sort( scores(~isnan(scores)) );
p = text2(sort_idx);

[ai ping Ng]

Since I have too many scores with NaN values which will affect my result, so I exclude scores with NaN values first and sort. However, by using yr method p = text2(sort_idx); , this will just consider for sorted_score location which had exclude NaN location. So, the location will not be the right location as what I want as I need to consider the NaN location also.

[Walter Roberson]

non_nan = find(~isnan(scores));
[sorted_score, sort_idx] = sort( scores(non_nan) );
p = text2( non_nan(sort_idx) );

[ai ping Ng]

Thanks, work well.