sklearn调参（验证曲线，可视化不同参数下交叉验证得分）

 一 、 原始方法：

思路：

1. 参数从 0+∞ 的一个 区间 取点， 方法如： np.logspace(-10, 0, 10) ， np.logspace(-6, -1, 5)

2. 循环调用cross_val_score计算得分。

在SVM不同的惩罚参数C下的模型准确率。

import matplotlib.pyplot as plt
from sklearn.model_selection import cross_val_score
import numpy as np
from sklearn import datasets, svm
digits = datasets.load_digits()
x = digits.data
y = digits.target
vsc = svm.SVC(kernel='linear')  

if __name__=='__main__':
    c_S = np.logspace(-10, 0, 10)#在范围内取是个对数
    # print ("length", len(c_S))
    scores = list()
    scores_std = list()
    for c in c_S:
        vsc.C = c
        this_scores = cross_val_score(vsc, x, y, n_jobs=4)#多线程 n_jobs，默认三次交叉验证
        scores.append(np.mean(this_scores))
        scores_std.append(np.std(this_scores))
    plt.figure(1, figsize=(4, 3))#绘图
    plt.clf()
    plt.semilogx(c_S, scores)#划线
    plt.semilogx(c_S, np.array(scores)+np.array(scores_std), 'b--')
    plt.semilogx(c_S, np.array(scores)-np.array(scores_std), 'b--')
    locs, labels = plt.yticks()
    plt.yticks(locs, list(map(lambda X: "%g" % X, locs)))#阶段点
    plt.ylabel('CV score')
    plt.xlabel('parameter C')
    plt.ylim(0, 1.1)#范围
    plt.show()  

效果：

二、高级方法（validation_curve）

思路：

直接用validation_curve获得模型在不同参数下，每次训练得分和测试得分。

from sklearn import svm
from sklearn.model_selection import validation_curve
from sklearn.datasets import load_digits
import numpy as np
import matplotlib.pyplot as plt
digits = load_digits()
X = digits.data
y = digits.target
param_range = np.logspace(-6, -1, 5)
vsc = svm.SVC()
train_score, test_score = validation_curve(vsc, X, y, param_name='gamma', param_range=param_range, cv=10, scoring="accuracy", n_jobs=1)
train_score_mean = np.mean(train_score, axis=1)
train_score_std = np.std(train_score, axis=1)
test_score_mean = np.mean(test_score, axis=1)
test_score_std = np.std(test_score, axis=1)
plt.title("validation curve with SVM")
plt.xlabel("$\gamma%")
plt.ylabel("Score")
plt.ylim()
lw = 2
plt.semilogx(param_range, train_score_mean,label="training score", color="darkorange", lw=lw)
plt.fill_between(param_range, train_score_mean-train_score_std, train_score_mean+train_score_std, alpha=0.2, color="navy", lw=lw)

plt.semilogx(param_range, test_score_mean,label="test score", color="blue", lw=lw)
plt.fill_between(param_range, test_score_mean-test_score_std, test_score_mean+test_score_std, alpha=0.2, color="navy", lw=lw)

plt.legend(loc="best")
plt.show()

结果：