def handle_iif(expr, split_expr, dictionary):
""" create the 4 rules related to the iif expression """
symb = "=>"
not_cdt = "!(%s)" % split_expr[0]
not_cc = "!(%s)" % split_expr[2]
rules = []
rules.append(Rule(expr, [split_expr[0], symb, split_expr[2]], dictionary))
rules.append(Rule(expr, [not_cdt, symb, not_cc], dictionary))
rules.append(Rule(expr, [split_expr[2], symb, split_expr[0]], dictionary))
rules.append(Rule(expr, [not_cc, symb, not_cdt], dictionary))
return rules
def __init__(self, text: str, number: int):
self.text: str = text.rstrip()
if '\n' in self.text:
raise GeneratorError(_MULTILINE_STRING_ERROR, number)
self.number: int = number
self.rules: list[Rule] = Rule.extract(number, text)
def evolve(self, epoch=1500):
while (self._generation < epoch):
for individual in self._population:
self._evaluate_fitness(individual)
self._display_population_fitness(self._population)
self._population = self._generate_offspring(self._population)
best = self._get_best_individual(self._population)
if (best.fitness == self._max_fitness):
break
self._reset_population_fitness(self._population)
if (self._generation != 0 and (self._generation + 1) % 5 == 0):
self.evalute_test(self._population)
self._reset_population_fitness(self._population)
self._generation += 1
rules = Rule.generate_rules_from_chromosome(best.chromosome,
self._feature_size,
self._rule_size)
self.evalute_test(self._population)
print("[BEST RULE BASE]\n")
for rule in rules:
print("RULE: {} {}".format(rule.feature, rule.label))
draw_graph(self._generation_info, epoch, self._crossover_probability,
self._mutation_probabilty, self._rule_count,
self._population_size, len(self._test_data))
self.test_rules(rules)
self.floating_point_to_rules(rules)
def _evaluate_fitness(self, individual):
rules = Rule.generate_rules_from_chromosome(individual.chromosome,
self._feature_size,
self._rule_size)
for data in self._data:
for rule in rules:
if (self._does_rule_match_data(rule.feature, data.feature)):
if (rule.label == int(data.label)):
individual.fitness += 1
break
def _evaluate_fitness(self, individual, use_test=False):
rules = Rule.generate_rules_from_chromosome(individual.chromosome,
self._feature_size,
self._rule_size)
dataset = self._test_data if use_test else self._train_data
for data in dataset:
for rule in rules:
if (self._does_rule_match(data.features, rule.feature)):
if (data.prediction == rule.label):
individual.fitness += 1
break
def create_rule(expr, dictionary):
""" create a list of Rule object with all the rules needed to resolve the
given expression
"""
split = split_expr(expr)
if None in split:
error(-1)
if '<' in split[1]:
rules = handle_iif(expr, split, dictionary)
else:
rules = [Rule(expr, split, dictionary)]
if td.Error in rules:
error(-1)
return rules
def evolve(self, epoch=1000):
while (self._generation < epoch):
self.__evaluate_population_fitness(self._population)
self._display_population_fitness(self._population)
self._population = self._generate_offspring(self._population)
best = self._get_best_individual(self._population)
if (best.fitness == self._max_fitness):
break
self._reset_population_fitness(self._population)
self._generation += 1
rules = Rule.generate_rules_from_chromosome(best.chromosome,
self._feature_size,
self._rule_size)
print("[BEST RULE BASE]")
for rule in rules:
rule.feature.append(rule.label)
print("{}".format(rule.feature))
draw_graph(self._generation_info, epoch, "Adaptive", "Adaptive",
self._rule_count, self._population_size)
def evolve(self, epochs=2000):
while (self._generation < epochs):
for individual in self._population:
self._evaluate_fitness(individual)
self._display_population_fitness(self._population)
self._population = self._generate_offspring(self._population)
best = self._get_best_individual(self._population)
if (best.fitness == self._max_fitness):
break
self._reset_population_fitness(self._population)
self._generation += 1
rules = Rule.generate_rules_from_chromosome(best.chromosome,
self._feature_size,
self._rule_size)
print("[BEST RULE BASE]")
for rule in rules:
print("RULE: {} {}".format(rule.feature, rule.label))
draw_graph(self._generation_info, epochs, self._crossover_probability,
self._mutation_probabilty, self._rule_count,
self._population_size)