def select_variables(size=DataManager.shared().get_number_of_columns()):
number_of_variables = randint(MIN_NUMBER_OF_VARIABLES,
MAX_NUMBER_OF_VARIABLES)
variables_to_use = sample(range(0, size), number_of_variables)
variables = []
for i in range(0, size):
if i in variables_to_use:
variables.append(1)
else:
variables.append(0)
return variables
def setup_population():
dm = DataManager.shared()
from genetic_algorithm.population import Population
from neural_network.neural_network import NeuralNetwork
population = Population.instantiate(NeuralNetwork.instantiate)
population.calculate_fitness(dm.get_X_train(), dm.get_y_train())
population.get_chromosomes().sort(key=lambda ann: ann.get_fitness(),
reverse=True)
print_population(population, 0)
return population
def main():
setup_data_manager()
dm = DataManager.shared()
from genetic_algorithm.genetic_algorithm import GeneticAlgorithm
population = setup_population()
generation_number = 1
while generation_number != NUMBER_OF_GENERATIONS and population.get_chromosomes(
)[0].get_fitness() < TARGET_FITNESS:
population = GeneticAlgorithm.evolve(population)
population.calculate_fitness(dm.get_X_train(), dm.get_y_train())
population.get_chromosomes().sort(key=lambda ann: ann.get_fitness(),
reverse=True)
print_population(population, generation_number)
generation_number += 1
def setup_data_manager():
data_manager = DataManager.shared()
data_manager.read_data(FILE_PATH)
# - Data preprocessing -
# Scaling
data_manager.scale_data(SCALE_EXCLUDE_VARIABLES)
# Encoding
data_manager.one_hot_encode_data(VARIABLES_TO_OH_ENCODE)
data_manager.binary_encode_data(VARIABLES_TO_B_ENCODE)
# Remove variables
data_manager.drop_columns(VARIABLES_TO_DELETE)
# - Data Split -
data_manager.split_data(VARIABLE_TO_CLASSIFY)
# - Data Analysis -
# Recursive feature elimination
data_manager.rfe_analysis(12)
# PCA analysis
data_manager.pca_analysis(0.95)
# - Print Variables -
data_manager.print_variables()