How to group different values in an array?

I have a test column array which contains 4 unique values? I have attached the array.

    intensityofcostcomposition
    __________________________

        "10.0,64.0,NaN"       
        "15.0,160.0,290.0"    
        "15.0,160.0,320.0"    
        "15.0,40.0,290.0"     

I want to create a new table column of same length as 'composition' array which will contain 3 different unique entries ('type1','type2','type3') . I want to group of "15.0,160.0,290.0" and "15.0,160.0,320.0" as 'type2'. How do I do it?

Answer 1

[Kevin Holly]

t = table;
t.intensityofcostcomposition = ["10.0,64.0,NaN"; 
        "15.0,160.0,290.0";
        "15.0,160.0,320.0";
        "15.0,40.0,290.0"]

t = 4×1 table
    intensityofcostcomposition
    __________________________

        "10.0,64.0,NaN"       
        "15.0,160.0,290.0"    
        "15.0,160.0,320.0"    
        "15.0,40.0,290.0"     

t.NewTableColumn = ['type1';'type2';'type2';'type3']

t = 4×2 table
    intensityofcostcomposition    NewTableColumn
    __________________________    ______________

        "10.0,64.0,NaN"               type1     
        "15.0,160.0,290.0"            type2     
        "15.0,160.0,320.0"            type2     
        "15.0,40.0,290.0"             type3     

Edit: with the attached .mat file:

load('composition.mat', 'ans')
t = ans;
t.new = t.intensityofcostcomposition
t.new(t.intensityofcostcomposition=="10.0,64.0,NaN")='type1'
t.new(t.intensityofcostcomposition=="15.0,40.0,290.0")='type3'
t.new(t.intensityofcostcomposition=="15.0,160.0,290.0")='type2'
t.new(t.intensityofcostcomposition=="15.0,160.0,320.0")='type2'

[Atanu]

Thank you!

Answer 2

[dpb]

Why are you holding numeric data as strings??? Even if stuff goes together somehow, use an array although it would seem likely given the variable name that the various components have to do with some particular measure of the cost and would probably be better off as variables reflecting such.

tT=array2table(str2double(split(C,',')),'VariableNames',{'C1','C2','C3'});  % turn into variables
[g,id1,id2]=findgroups(tT.C1,tT.C2);                                        % get group by first two components
ug=unique(g);                                                               % unique groups/types
tT.Class=categorical(g,ug,compose('Type%d',ug-1));                          % add class type

Above yields

>> tT
tT =
  4×4 table
    C1    C2     C3     Class
    __    ___    ___    _____
    10     64    NaN    Type0
    15    160    290    Type2
    15    160    320    Type2
    15     40    290    Type1
>> 

where I went ahead and aribtrarily also assigned a Class ID to the other elements a well; one can always simply turn those into <ismissing> or create another categorical value for them; whatever fits the problem.

To do something similar with the data still held as strings would require string-matching for the specific pattern; gets painful very quickly that way.

You need to be able to define the rules by which the Type code is to be assigned -- the rule inferred from the description of the desired output above is that the first two C are to be the same -- that works here but might not in general, depending upon just what the logic is that decides those are, indeed, Type 2.