Suppose I run a 7/11, and the following 100x3
cell array sorted by the first column, time, is my sale records.
12:32:01 customer1 12
12:32:02 customer2 13
12:32:04 customer6 4
12:32:06 customer8 6
12:32:07 customer1 9
12:32:07 customer1 6
12:32:12 customer2 1
...
As you have noticed, each customer can make purchases multiple times. e.g. customer 1 actually made three different payments.
I now wish to compute the average payment of each customer. e.g. let's assume customer 1 have only made 3 payments as shown above. Then, his average payment would be (12+9+6)/3=9
.
I can surely write a for loop to loop through all the entries and keep tracks of each customer. However, I feel that is not how it is supposed to be done with MATLAB.
So what is the most MATLAB-ish way of completing the task?
Start with unique
to get an integer "key" for each customer, then feed that into accumarray
with the @mean
function handle:
data = {'12:32:01','customer1',12; '12:32:02','customer2',13;...
'12:32:04','customer6',4; '12:32:06','customer8',6;...
'12:32:07','customer1',9; '12:32:07','customer1',6;...
'12:32:12','customer2',1};
[customers,~,ic] = unique(data(:,2));
avePayment = accumarray(ic,[data{:,3}],[],@mean);
Then assemble the outputs:
>> custAvgTab = [customers num2cell(avePayment)]
custAvgTab =
'customer1' [9]
'customer2' [7]
'customer6' [4]
'customer8' [6]
IMHO, this is quite MATLAB-ish and actually surprisingly intuitive.
NOTE: I replaced cell2mat(data(:,3))
with [data{:,3}]
since I think it is better to use built-in MATLAB operations when possible.
NOTE 2: For large data, sprintfc('%d',avePayment)
may be faster than num2cell(avePayment)
.