【量子AI #20】古典ML vs QML ベンチマーク比較の実装

8. パフォーマンスベンチマーク

8.1 古典MLとQMLの比較実験

import time, numpy as np, pandas as pd
from sklearn.svm import SVC
from sklearn.ensemble import RandomForestClassifier, GradientBoostingClassifier
from sklearn.neural_network import MLPClassifier
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import cross_val_score

class QMLBenchmark:
    def __init__(self, n_features=32, n_samples=500):
        self.n_features = n_features; self.n_samples = n_samples

    def benchmark_classical(self, X, y):
        scaler = StandardScaler()
        Xs = scaler.fit_transform(X)
        models = {
            "SVM (RBF)": SVC(kernel='rbf', probability=True, class_weight='balanced'),
            "Random Forest": RandomForestClassifier(n_estimators=100, class_weight='balanced'),
            "Gradient Boosting": GradientBoostingClassifier(n_estimators=100),
            "MLP (Classical NN)": MLPClassifier(hidden_layer_sizes=(64,32,16), max_iter=500)
        }
        results = {}
        for name, model in models.items():
            t0 = time.perf_counter()
            scores = cross_val_score(model, Xs, y, cv=5, scoring='roc_auc')
            elapsed = time.perf_counter() - t0
            results[name] = {"roc_auc": scores.mean(), "std": scores.std(), "time": elapsed}
            print(f"{name:25s}: ROC-AUC={scores.mean():.4f}+/-{scores.std():.4f} ({elapsed:.1f}s)")
        return results

benchmark = QMLBenchmark(n_features=32, n_samples=500)
pipeline = ManufacturingQMLPipeline(n_features=32)
X, y = pipeline.generate_synthetic_sensor_data(n_samples=500)
print(f"データセット: {X.shape} | 不良率: {y.mean():.1%}")
results = benchmark.benchmark_classical(X, y)