How I categorize a features?

1 回表示 (過去 30 日間)
HelpAStudent
HelpAStudent 2022 年 5 月 14 日
コメント済み: the cyclist 2022 年 5 月 14 日
Hi! I have a dataset like the histogram here: with some data around 0, some other around 1, 2, 3, 4 and 5.
I would like to make the features categorical as the amount at witch are they roughly equal in value.
This is the histogram of the features:
Please help me
  1 件のコメント
Image Analyst
Image Analyst 2022 年 5 月 14 日
It may or may not be possible. How were those data values determined?

サインインしてコメントする。

採用された回答

the cyclist
the cyclist 2022 年 5 月 14 日
編集済み: the cyclist 2022 年 5 月 14 日
Do you mean that you have numerical values, and you want to treat those as categorical instead? You can convert numeric to categorical using the categorical function.
x = 1:5
x = 1×5
1 2 3 4 5
c = categorical(x)
c = 1×5 categorical array
1 2 3 4 5
You said "roughly" equal in value, so maybe you need to do some rounding first?
x = [1.1 2.2 2.9 3.8 5.1]
x = 1×5
1.1000 2.2000 2.9000 3.8000 5.1000
c = categorical(round(x))
c = 1×5 categorical array
1 2 3 4 5
  1 件のコメント
the cyclist
the cyclist 2022 年 5 月 14 日
When I wrote this answer, I hadn't noticed that your values are not 1,2,3,4,5, but rather 10^-3 times that. So, you'll need to round differently:
x = [1.1 2.2 2.9 3.8 5.1]*1.e-3
x = 1×5
0.0011 0.0022 0.0029 0.0038 0.0051
rx = round(x,3)
rx = 1×5
0.0010 0.0020 0.0030 0.0040 0.0050
c = categorical(rx)
c = 1×5 categorical array
0.001 0.002 0.003 0.004 0.005

サインインしてコメントする。

その他の回答 (1 件)

Image Analyst
Image Analyst 2022 年 5 月 14 日
You can add a tiny bit of noise then recompute the histogram edges such that the bins will be equal percentages (heights). Like this:
data = [zeros(1, 1580), ones(1, 50), 2*ones(1, 70), 2*ones(1, 50), 3*ones(1, 40), 4*ones(1, 25), 4.7*ones(1, 10)]/1000;
subplot(2, 1, 1);
[counts, edges] = histcounts(data);
bar(edges(1:end-1), counts);
grid on;
title('Uneven Bars', 'FontSize', 20);
% Now add a tiny bit of noise and sort
noisyData = data + 0.000001 * rand(size(data));
sortedData = sort(noisyData);
% Get cdf
c = cumsum(sortedData);
c = rescale(c, 0, 100); % Convert to percent.
% Find 6 bins
numBins = 6;
indexes = round(linspace(1, length(data), numBins+1))
indexes = 1×7
1 305 609 913 1217 1521 1825
edges2 = sortedData(indexes)
edges2 = 1×7
0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0047
subplot(2, 1, 2);
counts2 = histcounts(noisyData, edges2)
counts2 = 1×6
304 304 304 304 304 305
bar(edges2(1:end-1), counts2);
grid on;
title('Even Bars', 'FontSize', 20);
  1 件のコメント
the cyclist
the cyclist 2022 年 5 月 14 日
I'll point out here that @Image Analyst seems to have interpreted your phrase "as the amount at witch are they roughly equal in value" to mean you want the bar heights to be equal.
I interpreted that differently, and took it to to mean that you wanted your data values to be equal (rather than "roughly equal"), which is why our two approaches are very different.

サインインしてコメントする。

カテゴリ

Help Center および File ExchangeCategorical Arrays についてさらに検索

製品


リリース

R2021a

Community Treasure Hunt

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

Start Hunting!

Translated by