Let’s say I have and array.shape = (296,3)

Where the first collumn contais 0 or 1, the second collumn contains 0,1 or 2 and the final collumn contains also 0,1 or 2.

I want to know how I can calculate all the probabilites for each one of the 18 possible combinations (2x3x3) of the 3 collumns. Possible sequences are [0,0,0];[0,0,1];etc… .

>Solution :

You can use numpy.unique to collect the unique sequences, along with their counts.

>>> (unique, counts) = numpy.unique(data, return_counts=True, axis=0)
>>> unique
array([0, 1, 1],
      [0, 1, 2],
      [1, 1, 1],
       ...])
>>> counts
array([2, 2, 1, ...])

Therefore the probability of the i-th element from uniques is its corresponding value from counts over the number of rows from your original data. In other words for example a count of 28 over 296 rows would be a probability of ~9.46%.