경기도미래기술학교 AI개발자 부트캠프 36일차 TIL- KNN 알고리즘 디시전 바운더리까지.

import numpy as np
import matplotlib.pyplot as plt

# np.random.seed(22)

K = 5
n_classes = 4
n_data = 50

X, y = [], []
for class_idx in range(n_classes):
    centroid = np.random.uniform(low=-10, high=10, size=(2,))
    X_ = np.random.normal(loc=centroid, scale=1.7, size=(n_data, 2))
    y_ = np.ones(n_data,) * class_idx

    X.append(X_); y.append(y_)

X = np.vstack(X)
y = np.concatenate(y)
# print(X.shape, y.shape)​
test_data = np.random.uniform(low=-10, high=10, size=(2,))

e_dists = np.sqrt(np.sum((X - test_data)**2, axis=1))
# print(e_dists.shape)

k_nearest_indices = np.argsort(e_dists)[:K]
# print(k_nearest_indices)
k_nearest_data = X[k_nearest_indices]
# print(k_nearest_data)​
k_nearest_y = y[k_nearest_indices]
print(k_nearest_y)

uniques, cnts = np.unique(k_nearest_y, return_counts=True)
print(uniques, cnts)
pred = int(uniques[np.argmax(cnts)])
print(f"test data is classified as class number {pred}")

fig, ax = plt.subplots(figsize=(10, 10))

# ax.scatter(X[:, 0], X[:, 1], c=y)
# ax.scatter(test_data[0], test_data[1], c='cyan', marker='*', s=100)​
for class_idx in range(n_classes):
    X_ = X[y == class_idx]
    ax.scatter(X_[:, 0], X_[:, 1], alpha=0.5)

    ax.scatter(test_data[0], test_data[1], color=f'C{pred}', s=200, marker="*")

    k_nearest_data = X[k_nearest_indices]
    k_nearest_y = y[k_nearest_indices]
    for data, y_ in zip(k_nearest_data, k_nearest_y):
        ax.plot([test_data[0], data[0]], [test_data[1], data[1]], ls=':', lw=2,
                color=f'C{int(y_)}')

ax.text(test_data[0]+0.5, test_data[1], f"class {pred}", color='fuchsia',
        fontsize=20)

plt.show()

# x1_lim, x2_lim = ax.get_xlim(), ax.get_ylim()
# x1 = np.linspace(x1_lim[0], x1_lim[1], 100)
# x2 = np.linspace(x2_lim[0], x2_lim[1], 100)
# X1, X2 = np.meshgrid(x1, x2)
# X_db = np.hstack([X1.reshape(-1, 1), X2.reshape(-1, 1)])
#
# preds = []
#
# for x_db in X_db:
#     distances = np.sum((X - x_db) ** 2, axis=1)
#     distances_argsort = np.argsort(distances)
#     close_K_indices = distances_argsort[:K]
#     close_classes = y[close_K_indices]
#     unique, cnts = np.unique(close_classes, return_counts=True)
#     pred = unique[np.argmax(cnts)]
#     preds.append(pred)
# preds = np.array(preds)
#
# print(len(preds))
# # ax.scatter(X1, X2, c=preds, alpha=0.1)
#
# for class_idx in range(n_classes):
#     X_ = X_db[preds == class_idx]
#     ax.scatter(X_[:, 0], X_[:, 1], color=f'C{class_idx}', alpha=0.1)
# plt.show()

강사님께서 주신 예제코드가 있는데 어지럽고 헷갈려서 계속 해야 하는 부분이다.