Introduction

其實對於SVM調節超引數不需要這麼複雜，因為gamma可能更重要一點，固定C=1，手動調節gamma即可。此外，sklearn的網格搜尋極其的慢，下面的程式碼出來結果至少要半個多小時，如果有經驗根本不需要。對於有經驗的人來說或許看學習曲線就能知道調什麼引數。但是為什麼還要這麼做呢？可能是為了裝吧，或許更直觀一點，不需要老中醫式的隨便開點良藥，看看效果再換藥了！

PCA：主成分分析

Pipline: 管道機制、官網

GridSearch：官網

Method

下面給出修改後的程式碼，裡面都有註釋，直接拿回去慢慢調：

資料是sklearn自帶的，資料量不大，如果是比賽資料，根本沒法跑，太慢了！！！

官網例程：比較三種降維方法：PCA+NMF(非負矩陣分解)+KBest

#!/usr/bin/python
# -*- coding: utf-8 -*-
"""
=================================================================
Selecting dimensionality reduction with Pipeline and GridSearchCV
=================================================================

This example constructs a pipeline that does dimensionality
reduction followed by prediction with a support vector
classifier. It demonstrates the use of GridSearchCV and
Pipeline to optimize over different classes of estimators in a
single CV run -- unsupervised PCA and NMF dimensionality
reductions are compared to univariate feature selection during
the grid search.
"""
# Authors: Robert McGibbon, Joel Nothman

from __future__ import print_function, division

import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import load_digits
from sklearn.model_selection import GridSearchCV
from sklearn.pipeline import Pipeline
from sklearn.model_selection import StratifiedShuffleSplit#分層洗牌分割交叉驗證
from sklearn.svm import LinearSVC
from sklearn.decomposition import PCA, NMF
from sklearn.feature_selection import SelectKBest, chi2

digits = load_digits()

print(__doc__)

pipe = Pipeline([
    ('reduce_dim', PCA()),
    ('classify', LinearSVC())
])

N_FEATURES_OPTIONS = [2, 4, 8]
C_OPTIONS = [1, 10, 100, 1000]
param_grid = [
    {
        'reduce_dim': [PCA(iterated_power=7), NMF()],
        'reduce_dim__n_components': N_FEATURES_OPTIONS,
        'classify__C': C_OPTIONS
    },
    {
        'reduce_dim': [SelectKBest(chi2)],
        'reduce_dim__k': N_FEATURES_OPTIONS,
        'classify__C': C_OPTIONS
    },
]
reducer_labels = ['PCA', 'NMF', 'KBest(chi2)']
cv = StratifiedShuffleSplit(n_splits=10, test_size=0.2, random_state=42)
grid = GridSearchCV(pipe, cv=3, n_jobs=2, param_grid=param_grid)

grid.fit(digits.data, digits.target)

mean_scores = np.array(grid.cv_results_['mean_test_score'])
# scores are in the order of param_grid iteration, which is alphabetical
mean_scores = mean_scores.reshape(len(C_OPTIONS), -1, len(N_FEATURES_OPTIONS))
# select score for best C
mean_scores = mean_scores.max(axis=0)
bar_offsets = (np.arange(len(N_FEATURES_OPTIONS)) *
               (len(reducer_labels) + 1) + .5)

plt.figure()
COLORS = 'bgrcmyk'
for i, (label, reducer_scores) in enumerate(zip(reducer_labels, mean_scores)):
    plt.bar(bar_offsets + i, reducer_scores, label=label, color=COLORS[i])

plt.title("Comparing feature reduction techniques")
plt.xlabel('Reduced number of features')
plt.xticks(bar_offsets + len(reducer_labels) / 2, N_FEATURES_OPTIONS)
plt.ylabel('Digit classification accuracy')
plt.ylim((0, 1))
plt.legend(loc='upper left')
plt.show()
 

尋找最優超引數：

# -*- coding: utf-8 -*-
"""
Created on Wed Jul 26 22:06:34 2017

@author: qiu
"""

from __future__ import print_function, division

import numpy as np
import matplotlib.pyplot as plt
from sklearn.datasets import load_digits
from sklearn.model_selection import GridSearchCV
from sklearn.pipeline import Pipeline
from sklearn.model_selection import StratifiedShuffleSplit#分層洗牌分割交叉驗證
from sklearn.svm import SVC
from sklearn.decomposition import PCA, NMF
from sklearn.feature_selection import SelectKBest, chi2

digits = load_digits()

#網格搜尋視覺化——熱力圖
pipe = Pipeline(steps=[
                       
    ('classify',  SVC())
])
C_range = np.logspace(-2, 1, 4)# logspace(a,b,N)把10的a次方到10的b次方區間分成N份
gamma_range = np.logspace(-9, -6, 4)
param_grid = [
    {
        'classify__C': C_range,
        'classify__gamma': gamma_range
    },
]

cv = StratifiedShuffleSplit(n_splits=10, test_size=0.2, random_state=42)

grid = GridSearchCV(pipe, param_grid=param_grid, cv=cv)#基於交叉驗證的網格搜尋。

grid.fit(digits.data, digits.target)

print("The best parameters are %s with a score of %0.2f"
      % (grid.best_params_, grid.best_score_))#找到最佳超引數
 

在獲取最佳引數後畫學習曲線，參考