sklearn之學習曲線

'''
    學習曲線：模型性能 = f(訓練集大小)
    學習曲線所需API：
            _, train_scores, test_scores = ms.learning_curve(
                                model,        # 模型
                                輸入集, 輸出集,
                                [0.9, 0.8, 0.7],    # 訓練集大小序列
                                cv=5        # 折疊數
                                )

    案例：在小汽車評級案例中使用學習曲線選擇訓練集大小最優參數。
'''

import numpy as np
import matplotlib.pyplot as mp
import sklearn.preprocessing as sp
import sklearn.ensemble as se
import sklearn.model_selection as ms
import sklearn.metrics as sm
import warnings

warnings.filterwarnings('ignore')

data = []
with open('./ml_data/car.txt', 'r') as f:
    for line in f.readlines():
        sample = line[:-1].split(',')
        data.append(sample)
data = np.array(data)
# print(data.shape)

# 整理好每一列的標簽編碼器encoders
# 整理好訓練輸入集與輸出集
data = data.T
# print(data.shape)
encoders = []
train_x, train_y = [], []
for row in range(len(data)):
    encoder = sp.LabelEncoder()
    if row < len(data) - 1:  # 不是最后列
        train_x.append(encoder.fit_transform(data[row]))
    else:  # 是最后一列，作為輸出集
        train_y = encoder.fit_transform(data[row])
    encoders.append(encoder)

train_x = np.array(train_x).T
# 訓練隨機森林分類器
model = se.RandomForestClassifier(max_depth=6, n_estimators=150, random_state=7)

# 繪制學習曲線
train_sizes = np.linspace(0.1, 1, 10)
_, train_scores, test_scores = ms.learning_curve(model, train_x, train_y, train_sizes=train_sizes, cv=5)
print(test_scores)
print(np.mean(test_scores,axis=1))


# 訓練之前進行交叉驗證
cv = ms.cross_val_score(model, train_x, train_y, cv=4, scoring='f1_weighted')
print(cv.mean())
model.fit(train_x, train_y)

# 自定義測試集，預測小汽車的等級
# 保證每個特征使用的標簽編碼器與訓練時使用的標簽編碼器匹配
data = [
    ['high', 'med', '5more', '4', 'big', 'low', 'unacc'],
    ['high', 'high', '4', '4', 'med', 'med', 'acc'],
    ['low', 'low', '2', '4', 'small', 'high', 'good'],
    ['low', 'med', '3', '4', 'med', 'high', 'vgood']]

data = np.array(data).T
test_x, test_y = [], []
for row in range(len(data)):
    encoder = encoders[row]  # 每列對應的標簽編碼器
    if row < len(data) - 1:
        test_x.append(encoder.transform(data[row]))  # 這里需要訓練了，直接轉換
    else:
        test_y = encoder.transform(data[row])
test_x = np.array(test_x).T

pred_test_y = model.predict(test_x)
print(pred_test_y)
pred_test_y = encoders[-1].inverse_transform(pred_test_y)
test_y = encoders[-1].inverse_transform(test_y)
print(pred_test_y)
print(test_y)

# 畫圖顯示學習曲線
mp.figure('Learning Curve', facecolor='lightgray')
mp.title('Learning Curve')
mp.xlabel('train size')
mp.ylabel('f1 score')
mp.grid(linestyle=":")
mp.plot(train_sizes, np.mean(test_scores, axis=1), label='Learning Curve')
mp.legend()

mp.show()


輸出結果：

[[0.69942197 0.69942197 0.69942197 0.69942197 0.70348837]
 [0.67630058 0.79768786 0.69942197 0.71965318 0.70348837]
 [0.66184971 0.70231214 0.75433526 0.74855491 0.70348837]
 [0.71098266 0.78323699 0.74277457 0.73988439 0.7005814 ]
 [0.71387283 0.71965318 0.5982659  0.74277457 0.74127907]
 [0.71387283 0.76878613 0.70809249 0.74855491 0.73837209]
 [0.71387283 0.7716763  0.72254335 0.82080925 0.75872093]
 [0.71387283 0.76878613 0.72254335 0.83526012 0.75872093]
 [0.71387283 0.7716763  0.73121387 0.83526012 0.76744186]
 [0.73121387 0.76878613 0.72254335 0.8583815  0.86046512]]
[0.70023525 0.71931039 0.71410808 0.735492   0.70316911 0.73553569
 0.75752453 0.75983667 0.763893   0.78827799]
0.7477732938195376
[2 0 0 3]
['unacc' 'acc' 'acc' 'vgood']
['unacc' 'acc' 'good' 'vgood']