MATLAB Answers


Converting categorical data to prespecified numbers

Danielle Leblance さんによって質問されました 2018 年 5 月 8 日
最新アクティビティ Walter Roberson
さんによって 編集されました 2018 年 5 月 8 日
I have a categorical array : ['SC' 'SC' 'SC' 'SC' 'SC' 'SC' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'MI' 'MI' 'MI' 'MI' 'MI' 'MI' 'MI' 'MI' 'MN' 'MN' 'MN' 'MN' 'MN' 'MN'];
I want to convert each category to a number of my choice: for instance SC should be 23, GA=10;MI=13,MN=15 etc... How can I do so?

  0 件のコメント

サインイン to comment.


2 件の回答

回答者: Guillaume
2018 年 5 月 8 日

keys = categorical({'SC', 'GA', 'MI', 'MN'});
values = [23, 10, 13, 15];
[found, where] = ismember(yourcategoricalarray, keys)
correspondingvalues = nan(size(yourcategoricalarray));
correspondingvalues(found) = values(where(found));
correspondingvalues will be nan for those entries you don't care about.

  0 件のコメント

サインイン to comment.

Ameer Hamza
回答者: Ameer Hamza
2018 年 5 月 8 日

You can do the do this as follow,
data = {'SC' 'SC' 'SC' 'SC' 'SC' 'SC' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA'...
'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA'...
'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA' 'GA'...
'GA' 'GA' 'GA' 'GA' 'MI' 'MI' 'MI' 'MI' 'MI' 'MI' 'MI' 'MI' 'MN' 'MN'...
'MN' 'MN' 'MN' 'MN'};
dataCategorical = categorical(data);
tablePattern = categorical({'SC', 'GA', 'MI', 'MN'}); % make list of all unique pattern
valueValue = [23, 10, 13, 15]; % make list of all values corrosponding to unique patterns
index = (dataCategorical' == tablePattern)*(1:4)';
data2Value = valueValue(index);

  6 件のコメント

Ameer Hamza
2018 年 5 月 8 日
What you want to do about those categories. Do you want to just remove them or make them equal to zero? If you just remove them then you will have no way of knowing that which element in data2Value corresponds to which element in data because several values are missing. It depend on what you want to do with those values.
Danielle Leblance 2018 年 5 月 8 日
make them nan. I will look at Guillaume's answer
2018 年 5 月 8 日
Note that you could have used the same method as in my answer (construct nan matrix, then fill with valueValue(index)) with Ameer's method. However using ismember is probably faster and certainly a lot less demanding in memory than the 2D array generated by the implicit expansion of ==

サインイン to comment.

Translated by