Source code for streamline.models.genetic_programming

from abc import ABC
from streamline.modeling.basemodel import BaseModel
from streamline.modeling.parameters import get_parameters
from gplearn.genetic import SymbolicClassifier as GP


[docs] class GPClassifier(BaseModel, ABC): model_name = "Genetic Programming" small_name = "GP" color = "purple" def __init__(self, cv_folds=3, scoring_metric='balanced_accuracy', metric_direction='maximize', random_state=None, cv=None, n_jobs=None): super().__init__(GP, "Genetic Programming", cv_folds, scoring_metric, metric_direction, random_state, cv) self.param_grid = get_parameters(self.model_name) self.param_grid['random_state'] = [random_state, ] self.small_name = "GP" self.color = "purple" self.n_jobs = n_jobs
[docs] def objective(self, trial, params=None): feature_names = params['feature_names'] self.params = {'population_size': trial.suggest_int('population_size', self.param_grid['population_size'][0], self.param_grid['population_size'][1]), 'generations': trial.suggest_int('generations', self.param_grid['generations'][0], self.param_grid['generations'][1]), 'tournament_size': trial.suggest_int('tournament_size', self.param_grid['tournament_size'][0], self.param_grid['tournament_size'][1]), 'function_set': trial.suggest_categorical('function_set', self.param_grid['function_set']), 'init_method': trial.suggest_categorical('init_method', self.param_grid['init_method']), 'parsimony_coefficient': trial.suggest_float('parsimony_coefficient', self.param_grid['parsimony_coefficient'][0], self.param_grid['parsimony_coefficient'][1]), 'feature_names': trial.suggest_categorical('feature_names', [feature_names]), 'low_memory': trial.suggest_categorical('low_memory', self.param_grid['low_memory']), 'random_state': trial.suggest_categorical('random_state', self.param_grid['random_state'])} mean_cv_score = self.hyper_eval() return mean_cv_score