ChatGPTを使って機械学習モデルを実装してみたい（遺伝的アルゴリズムによる変数選択編）

Human

遺伝的アルゴリズムを用いて、pythonで機械学習の説明変数選択を行うクラスを作成してください。　予測誤差の評価はcrossvalidationのr2を用い、回帰モデルはPLSRegressionを使ってください。　遺伝的アルゴリズムのライブラリはdeapを使ってください。　関数fitとtransformは必須です

ChatGPT

Here is a class implementation in Python that performs feature selection using a genetic algorithm, uses cross-validated R2 as the evaluation metric, and uses PLSRegression as the regression model:

class GeneticFeatureSelection:
    def __init__(self, population_size=100, generations=20, cxpb=0.5, mutpb=0.2):
        self.population_size = population_size
        self.generations = generations
        self.cxpb = cxpb
        self.mutpb = mutpb
        
    def fit(self, X, y):
        self.X = X
        self.y = y
        self.num_features = X.shape[1]
        
        # Create a fitness function to evaluate the feature sets
        def eval_features(individual):
            individual = np.array(individual) == 1
            if individual.sum() == 0:
                return 0,
            X_selected = X[:, individual]
            model = PLSRegression()
            model.fit(X_selected, y)
            r2 = model.score(X_selected, y)
            return r2,
        
        creator.create("FitnessMax", base.Fitness, weights=(1.0,))
        creator.create("Individual", list, fitness=creator.FitnessMax)
        
        toolbox = base.Toolbox()
        toolbox.register("attr_bool", np.random.randint, 0, 1)
        toolbox.register("individual", tools.initRepeat, creator.Individual, toolbox.attr_bool, n=self.num_features)
        toolbox.register("population", tools.initRepeat, list, toolbox.individual)
        toolbox.register("evaluate", eval_features)
        toolbox.register("mate", tools.cxTwoPoint)
        toolbox.register("mutate", tools.mutFlipBit, indpb=0.05)
        toolbox.register("select", tools.selTournament, tournsize=3)
        
        population = toolbox.population(n=self.population_size)
        self.hof = tools.HallOfFame(1)
        stats = tools.Statistics(lambda ind: ind.fitness.values)
        stats.register("avg", np.mean)
        stats.register("std", np.std)
        stats.register("min", np.min)
        stats.register("max", np.max)
        
        population, log = algorithms.eaSimple(population, toolbox, cxpb=self.cxpb, mutpb=self.mutpb, ngen=self.generations, 
                                             stats=stats, halloffame=self.hof, verbose=True)
        
    def transform(self, X):
        selected_features = np.array(self.hof[0]) == 1
        return X[:, selected_features]

基本的なコードは出来ているので、後は修正すれば使えそう