Linear Discriminant Analysis (LDA) | หาแนวที่แยกคลาสได้ดีที่สุด

from __future__ import annotations

import japanize_matplotlib
import matplotlib.pyplot as plt
import numpy as np
from sklearn.datasets import make_blobs
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.metrics import accuracy_score


def run_lda_demo(
    n_samples: int = 200,
    random_state: int = 42,
    title_boundary: str = "เส้นแบ่งของ LDA",
    title_projection: str = "การฉายด้วย LDA",
    xlabel: str = "คุณลักษณะที่ 1",
    ylabel: str = "คุณลักษณะที่ 2",
    hist_xlabel: str = "ค่าหลังฉาย",
    class0_label: str = "คลาส 0",
    class1_label: str = "คลาส 1",
) -> dict[str, float]:
    """Train LDA on synthetic blobs and plot boundary plus projection."""
    japanize_matplotlib.japanize()
    X, y = make_blobs(
        n_samples=n_samples,
        centers=2,
        n_features=2,
        cluster_std=2.0,
        random_state=random_state,
    )

    clf = LinearDiscriminantAnalysis(store_covariance=True)
    clf.fit(X, y)

    accuracy = float(accuracy_score(y, clf.predict(X)))
    w = clf.coef_[0]
    b = float(clf.intercept_[0])

    xs = np.linspace(X[:, 0].min() - 1, X[:, 0].max() + 1, 300)
    ys_boundary = -(w[0] / w[1]) * xs - b / w[1]

    fig, ax = plt.subplots(figsize=(7, 6))
    ax.set_title(title_boundary)
    ax.scatter(X[:, 0], X[:, 1], c=y, cmap="coolwarm", edgecolor="k", alpha=0.8)
    ax.plot(xs, ys_boundary, "k--", lw=1.2, label="w^T x + b = 0")
    ax.set_xlabel(xlabel)
    ax.set_ylabel(ylabel)
    ax.legend(loc="best")
    ax.grid(alpha=0.25)
    fig.tight_layout()
    plt.show()

    X_proj = clf.transform(X)[:, 0]

    fig, ax = plt.subplots(figsize=(8, 4))
    ax.set_title(title_projection)
    ax.hist(X_proj[y == 0], bins=20, alpha=0.7, label=class0_label)
    ax.hist(X_proj[y == 1], bins=20, alpha=0.7, label=class1_label)
    ax.set_xlabel(hist_xlabel)
    ax.legend(loc="best")
    ax.grid(alpha=0.25)
    fig.tight_layout()
    plt.show()

    return {"accuracy": accuracy}


metrics = run_lda_demo(
    title_boundary="เส้นแบ่งของ LDA",
    title_projection="การฉายด้วย LDA",
    xlabel="คุณลักษณะที่ 1",
    ylabel="คุณลักษณะที่ 2",
    hist_xlabel="ค่าหลังฉาย",
    class0_label="คลาส 0",
    class1_label="คลาส 1",
)
print(f"ความแม่นยำขณะฝึก: {metrics['accuracy']:.3f}")