Is it possible to predict random number with scikit-learn?

I have 2 data sets, one is sequence of numbers [1,54] and second one is connected with the first one but except numbers I have colors (gray, red, blue, gold), basically 0-gold, 1-blue, 2-gray, 3-red, 4-gray, 5-red and so on. There are Ni numbers of each color, 26 gray numbers, 17 red, 10 blue and 1 gold. Which means, just by this, I have 1,85*Ni percents to hit each color. But because machine is generating those numbers, I want to know if it is possible to predict (50%+ accuracy) next color/number with some probability for each color. Mainly I need color, but if it's going to be more precise with numbers I will go with it. Example: in my data set I know that after red-blue-red-gray I get gold 67% of time, 14% gray, 15% blue and 4% red. I want to use Python 2.7 and I planned to use scikit-learn module but if you can suggest me anything more efficient I would consider that, too. Thank you very much!