Source code for infpy.decision.test.genepy_test

#
# Copyright John Reid 2007, 2010
#


import unittest, logging
from infpy.genepy import Population
from infpy.decision import Context, OrdinalAttribute, EnumerativeAttribute, ContinuousAttribute, DecisionTreeSpecies, count_nodes, log_tree

def ord(data): return data[0]
def enum(data): return data[1]
def cont(data): return data[2]
class GenepyTest( unittest.TestCase ):
    """Test case for decision tree genepy interface"""

    # define the context - i.e. what attributes the data have and their
    # classification
    context = Context()
    context.attributes.append(OrdinalAttribute('ordinal', ord, 10))
    context.attributes.append(EnumerativeAttribute('enumerative', enum, 4))
    context.attributes.append(ContinuousAttribute('continuous', cont, 0.0, 1.0))
    context.outcomes = [ 0, 1, 2 ]

    # generate some data
    from random import gauss, randint
    data = [
      ((randint(0,9),randint(0,3),gauss(0.0,1.0)),randint(0,2))
      for i in xrange(30)
    ]

    species = DecisionTreeSpecies(context)
    species.data = data

    def fitness_fn(self, individual):
        loss = 0
        for x, y in self.data:
            if y != individual(x):
                loss += 1
        return loss

    def test_initialisation(self):
        self.species.random_individual()

    def test_mutation(self):
        pass #self.species.mutate(self.species.random_individual())

    def test_combination(self):
        self.species.mate(self.species.random_individual(), self.species.random_individual())

    def test_call(self):
        for x, _y in self.data: self.species.random_individual()(x)

    def test_overall(self):
        pop = Population(size = 100, species = self.species, fitness_fn = self.fitness_fn)
        pop.post_generation_process = self.species.prune_individuals
        for gen_idx in xrange(100):
            # execute a generation
            pop.generation()

            # and print how good best is
            tree_sizes = list(count_nodes(i) for i in pop.individuals)
            logging.debug("%3d; best fitness=%2d; average tree size: %3.3f", gen_idx, int(pop.most_fit.fitness), float(sum(tree_sizes))/len(tree_sizes))
        log_tree(pop.most_fit, logging.getLogger(), level=logging.DEBUG)

if __name__ == "__main__":
    logging.basicConfig(level=logging.DEBUG)
    unittest.main()