#%%導(dǎo)入基本庫(kù)
# 基礎(chǔ)數(shù)組運(yùn)算庫(kù)導(dǎo)入
import numpy as np 
# 畫圖庫(kù)導(dǎo)入
import matplotlib.pyplot as plt 
# 導(dǎo)入三維顯示工具
from mpl_toolkits.mplot3d import Axes3D
# 導(dǎo)入LDA模型
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
# 導(dǎo)入demo數(shù)據(jù)制作方法
from sklearn.datasets import make_classification
#%%模型訓(xùn)練
# 制作四個(gè)類別的數(shù)據(jù)，每個(gè)類別100個(gè)樣本
X, y = make_classification(n_samples=1000, n_features=3, n_redundant=0,
                           n_classes=4, n_informative=2, n_clusters_per_class=1,
                           class_sep=3, random_state=10)
# 將四個(gè)類別的數(shù)據(jù)進(jìn)行三維顯示
fig = plt.figure()
ax = Axes3D(fig, rect=[0, 0, 1, 1], elev=20, azim=20)
ax.scatter(X[:, 0], X[:, 1], X[:, 2], marker='o', c=y)
plt.show()

#%%建立 LDA 模型
lda = LinearDiscriminantAnalysis()
# 進(jìn)行模型訓(xùn)練
lda.fit(X, y)
#%%查看lda的參數(shù)
print(lda.get_params())

#%%數(shù)據(jù)可視化
#模型預(yù)測(cè)
X_new = lda.transform(X)
# 可視化預(yù)測(cè)數(shù)據(jù)
plt.scatter(X_new[:, 0], X_new[:, 1], marker='o', c=y)
plt.show()

#%%使用新的數(shù)據(jù)進(jìn)行測(cè)試
a = np.array([[-1, 0.1, 0.1]])
print(f"{a} 類別是: ", lda.predict(a))
print(f"{a} 類別概率分別是: ", lda.predict_proba(a))

a = np.array([[-12, -100, -91]])
print(f"{a} 類別是: ", lda.predict(a))
print(f"{a} 類別概率分別是: ", lda.predict_proba(a))

a = np.array([[-12, -0.1, -0.1]])
print(f"{a} 類別是: ", lda.predict(a))
print(f"{a} 類別概率分別是: ", lda.predict_proba(a))

a = np.array([[0.1, 90.1, 9.1]])
print(f"{a} 類別是: ", lda.predict(a))
print(f"{a} 類別概率分別是: ", lda.predict_proba(a))

#%%導(dǎo)入庫(kù)函數(shù)
# 導(dǎo)入手寫數(shù)據(jù)集 MNIST
from sklearn.datasets import load_digits
# 導(dǎo)入訓(xùn)練集分割方法
from sklearn.model_selection import train_test_split
# 導(dǎo)入LDA模型
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
# 導(dǎo)入預(yù)測(cè)指標(biāo)計(jì)算函數(shù)和混淆矩陣計(jì)算函數(shù)
from sklearn.metrics import classification_report, confusion_matrix
# 導(dǎo)入繪圖包
import seaborn as sns
import matplotlib
import matplotlib.pyplot as plt
#%% 導(dǎo)入MNIST數(shù)據(jù)集
mnist = load_digits()
# 查看數(shù)據(jù)集信息
print('The Mnist dataeset:\n',mnist)

# 分割數(shù)據(jù)為訓(xùn)練集和測(cè)試集
x, test_x, y, test_y = train_test_split(mnist.data, mnist.target, test_size=0.1, random_state=2)

#%%## 輸出示例圖像
images = range(0,9)

plt.figure(dpi=100)
for i in images:
    plt.subplot(330 + 1 + i)
    plt.imshow(x[i].reshape(8, 8), cmap = matplotlib.cm.binary,interpolation="nearest")
# show the plot
plt.show()

#%%利用LDA對(duì)手寫數(shù)字進(jìn)行訓(xùn)練與預(yù)測(cè)
m_lda = LinearDiscriminantAnalysis()# 建立 LDA 模型
# 進(jìn)行模型訓(xùn)練
m_lda.fit(x, y)
# 進(jìn)行模型預(yù)測(cè)
x_new = m_lda.transform(x)
# 可視化預(yù)測(cè)數(shù)據(jù)
plt.scatter(x_new[:, 0], x_new[:, 1], marker='o', c=y)
plt.title('MNIST with LDA Model')
plt.show()

#%% 進(jìn)行測(cè)試集數(shù)據(jù)的類別預(yù)測(cè)
y_test_pred = m_lda.predict(test_x)
print("測(cè)試集的真實(shí)標(biāo)簽:\n", test_y)
print("測(cè)試集的預(yù)測(cè)標(biāo)簽:\n", y_test_pred)
#%% 進(jìn)行預(yù)測(cè)結(jié)果指標(biāo)統(tǒng)計(jì) 統(tǒng)計(jì)每一類別的預(yù)測(cè)準(zhǔn)確率、召回率、F1分?jǐn)?shù)
print(classification_report(test_y, y_test_pred))
# 計(jì)算混淆矩陣
C2 = confusion_matrix(test_y, y_test_pred)
# 打混淆矩陣
print(C2)

# 將混淆矩陣以熱力圖的防線顯示
sns.set()
f, ax = plt.subplots()
# 畫熱力圖
sns.heatmap(C2, cmap="YlGnBu_r", annot=True, ax=ax)  
# 標(biāo)題 
ax.set_title('confusion matrix')
# x軸為預(yù)測(cè)類別
ax.set_xlabel('predict')  
# y軸實(shí)際類別
ax.set_ylabel('true')  
plt.show()