Saving results from loop with intervals

Hi everyone, I am applying Naive Bayes Classification by hand to a data set with formula. My data has 45.000 columns and no class informations. I would like to seperate it with classes. I will have nearly 100 classes but I don't want to write it by hand. I want to handle it with loop and I am looking for hours but I couldn't figure it out.

veri = readtable("2018.csv");

maxgh = max(veri.GlobalHoriz_W_m_2_);

mingh = min(veri.GlobalHoriz_W_m_2_);

meangh = mean(veri.GlobalHoriz_W_m_2_);

stdgh = std(veri.GlobalHoriz_W_m_2_);

vericell = table2cell(veri);

ghcolumn = cell2mat(vericell(:,3));

*class1 = vericell(ghcolumn>=mingh & ghcolumn <=95,:);

class2 = vericell(ghcolumn>95 & ghcolumn <=195,:);

class3 = vericell(ghcolumn>195 & ghcolumn <=295,:);

class4 = vericell(ghcolumn>295 & ghcolumn <=395,:);

class5 = vericell(ghcolumn>395 & ghcolumn <=495,:);

class6 = vericell(ghcolumn>495 & ghcolumn <=595,:);

class7 = vericell(ghcolumn>595 & ghcolumn <=695,:);

class8 = vericell(ghcolumn>695 & ghcolumn <=795,:);

class9 = vericell(ghcolumn>795 & ghcolumn <=895,:);

class10 = vericell(ghcolumn>895 & ghcolumn <=995,:);

class11 = vericell(ghcolumn>995 & ghcolumn <=maxgh,:);*

I would like to create seperate tables for the values which is between -5,95 95,195 195,295 etc.. with loop of course

For now I increased the numbers for 100 but I would like to increase them by 10 and I don't want to do it one by one with hand . Any help will be appreciated so much. Thanks in advance.

[Guillaume]

Your whole code shows that you don't know how to manipulate tables. For example, your inefficient table2cell + cell2mat is simply

ghcolumn = veri{:, 3};

Creating numbered variables, manually or by code, is always a bad idea. You should have a single array/cell array/whatever container instead of these numbered variables.

It is possible to split your tables into a cell array of tables with just a few lines of code (loop not needed). However, it is just as likely that you don't need to do that and that whatever you want to do afterward will be much easier if you don't split your table at all. Therefore, can you explain why you want to do it?

[Stephen23]

@Sukru Yavuz: in general you should keep data together as much as possible, because this usually makes processing it easier and means that you do not create unnecessary duplicates of the same data in memory (what a waste!).

Most likely the table class already has some grouping method that would help you to achieve your goal, in which case that would be a much simpler solution: so what do you actually want to do with this data?

Note that creating variable names like that in a loop is very inefficient, and is not recommended:

https://www.mathworks.com/matlabcentral/answers/304528-tutorial-why-variables-should-not-be-named-dynamically-eval

[Sukru Yavuz]

@Guillaume I have to create different classes to increase the accuracy of Naive Bayes formula. I will test a date to see which class it belongs to. That's why I am creating different class tables. And you know By narrowing the intervals the number of classes will increase and also accuracy will increase.

@Stephen Cobeldick After seperating it to different classes, I will select a date and find which class does that date belongs to. That's why I am creating different tables to be able to calculate the possibilities for each of them.

[Sukru Yavuz]

My advisor told me to do it this way but like you said im pretty sure there are more efficient ways than that. Let's just say I would like to group my class label. The minimum value of that column is -5.071, the maximum value of that is 1008. I would like to set the intervals as 20 and split them.

Answer 1

[Guillaume]

As Stephen and I said, splitting your table into multiple tables is probably going to complicate things for you. I would recommend you create a new column instead which tells you which class each row belongs to:

veri = readtable("2018.csv");
veri.Class = discretize(veri{:, 3}, min(veri{:, 3}):20:max(veri{:, 3})+20)

That's all that is needed. No loop required. You can then do calculations by class using rowfun or varfun with the 'GroupingVariable', 'Class' option. For example, to get the mean of column 4 per each class:

varfun(@mean, veri, 'InputVariables', 4, 'GroupingVariables', 'Class')

If you really want to split the table:

splittables = splitapply(@(rows) {veri(rows, :)}, (1:height(veri))', veri.Class)

splittables{i} is the table of class i. But again, this should not be needed. Calculating the above mean is more complicated once it's split.

[Sukru Yavuz]

Now it is making more sense. I understood it completely. Thank you for being this clear sir. Have a nice evening.