默认SVC训练模型，6个特征变量，训练集准确率：96.0%，测试集准确率：92.4%
默认SVC训练模型，10个特征变量，训练集准确率：98.7% ，测试集准确率：98.2%
LinearSVC训练模型， 6个特征变量， 训练集准确率：93.9%，测试集准确率：92.3%
LinearSVC训练模型， 10个特征变量， 训练集准确率：99.4%，测试集准确率：96.0%

结论：
1. 增加特征变量可以提高准确率，可能是因为模型维度变高，模型变得更加复杂。可以看出特征变量的选取很重要。
2. 训练集拟合都比较好，但是测试集准确率出现不同程度的下降。
3. 模型训练的准确率与人类水平之间偏差可以通过增加特征变量或采用新的训练模型来降低；模型训练的准确率与测试集测试的准确率之间的方差可以通过正则化，提高泛化性能等方式来降低。

# encoding=utf-8
from sklearn import svm
from sklearn import metrics
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

# 加载数据集，你需要把数据放到目录中
data = pd.read_csv("./data.csv")
# 数据探索
# 因为数据集中列比较多，我们需要把dataframe中的列全部显示出来
pd.set_option('display.max_columns', None)
#print(data.columns)
#print(data.head(5))
#print(data.describe())

# 将特征字段分成3组
features_mean= list(data.columns[2:12])
features_se= list(data.columns[12:22])
features_worst=list(data.columns[22:32])
# 数据清洗
# ID列没有用，删除该列
data.drop("id",axis=1,inplace=True)
# 将B良性替换为0，M恶性替换为1
data['diagnosis']=data['diagnosis'].map({'M': 1, 'B': 0})

# 特征选择
features_remain = ['radius_mean', 'texture_mean', 'perimeter_mean',
       'area_mean', 'smoothness_mean', 'compactness_mean', 'concavity_mean',
       'concave points_mean', 'symmetry_mean', 'fractal_dimension_mean',
       'radius_se', 'texture_se', 'perimeter_se', 'area_se', 'smoothness_se',
       'compactness_se', 'concavity_se', 'concave points_se', 'symmetry_se',
       'fractal_dimension_se', 'radius_worst', 'texture_worst',
       'perimeter_worst', 'area_worst', 'smoothness_worst',
       'compactness_worst', 'concavity_worst', 'concave points_worst',
       'symmetry_worst', 'fractal_dimension_worst']

# 抽取30%的数据作为测试集，其余作为训练集
train, test = train_test_split(data, test_size = 0.3)# in this our main data is splitted into train and test
# 抽取特征选择的数值作为训练和测试数据
train_X = train[features_remain]
train_y=train['diagnosis']
test_X= test[features_remain]
test_y =test['diagnosis']

# 采用Z-Score规范化数据，保证每个特征维度的数据均值为0，方差为1
ss = StandardScaler()
train_X = ss.fit_transform(train_X)
test_X = ss.transform(test_X)

# 创建SVM分类器
model = svm.LinearSVC()
# 用训练集做训练
model.fit(train_X,train_y)
# 用测试集做预测
prediction=model.predict(test_X)
print('准确率: ', metrics.accuracy_score(prediction,test_y))

准确率: 0.9707602339181286

选取全部特征：
SVM分类器准确率： 0.9824561403508771
cross_val_score的准确率为：0.9727
linearSVM分类器的准确率： 0.9766081871345029
cross_val_score的准确率为：0.9652

选取mean相关特征：
SVM分类器准确率： 0.9239766081871345
cross_val_score的准确率为：0.9321
linearSVM分类器的准确率： 0.9298245614035088
cross_val_score的准确率为：0.9247

数据结果上看：
SVM的结果要好于linearSVM;
选取多特征的结果要好于选取少特征的结果

采用linearSVC, 预测准确率更高。
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn import svm
from sklearn.metrics import accuracy_score
from sklearn.preprocessing import StandardScaler

data = pd.read_csv('data.csv')
data.drop('id', axis=1, inplace=True)

feature_names = list(data.columns)
feature_names.remove('diagnosis')

traindata, testdata = train_test_split(data, test_size=0.3)
train_x = traindata[feature_names]
train_y = traindata['diagnosis']
test_x = testdata[feature_names]
test_y = testdata['diagnosis']

ss = StandardScaler()
train_x = ss.fit_transform(train_x)
test_x = ss.transform(test_x)
                    
model = svm.LinearSVC()
model.fit(train_x, train_y)
prediction = model.predict(test_x)
print("The accuracy is %f" % accuracy_score(prediction, test_y))

语言Python3.6 没有z-score规范化数据以及规范化后两种情况前提预测准确率，使用LinearSVC，选取所有mean属性

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import train_test_split
from sklearn import svm
from sklearn import metrics
from sklearn.preprocessing import StandardScaler

#导入数据
path = '/Users/apple/Desktop/GitHubProject/Read mark/数据分析/geekTime/data/'
data = pd.read_csv(path + 'breast_cancer/data.csv')

#数据探索
pd.set_option('display.max_columns', None)
print(data.columns)
print(data.head(5))
print(data.describe())

#将特征字段进行分组
features_mean = list(data.columns[2:12])
features_se = list(data.columns[12:22])
features_worst = list(data.columns[22:32])

#数据清洗
#删除ID列
data.drop('id',axis=1,inplace=True)
#将良性B替换为0，将恶性替换为1
data['diagnosis'] = data['diagnosis'].map({'B':0,'M':1})

#将肿瘤诊断结果可视化
sns.countplot(data['diagnosis'],label='count')
plt.show()
#计算相关系数
corr = data[features_mean].corr()
plt.figure(figsize=(14,14))

#用热力图呈现相关性，显示每个方格的数据
sns.heatmap(corr,annot=True)
plt.show()

#特征选择，选择所有的mean数据
feature_remain = ['radius_mean', 'texture_mean', 'perimeter_mean',
       'area_mean', 'smoothness_mean', 'compactness_mean', 'concavity_mean',
       'concave points_mean', 'symmetry_mean', 'fractal_dimension_mean']

#抽取30%特征选择作为测试数据，其余作为训练集
train,test = train_test_split(data,test_size=0.3)
#抽取特征选择作为训练和测试数据
train_data = train[feature_remain]
train_result = train['diagnosis']
test_data = test[feature_remain]
test_result = test['diagnosis']

#创建SVM分类器
model = svm.LinearSVC()
#用训练集做训练
model.fit(train_data,train_result)
#用测试集做预测
prediction = model.predict(test_data)
#准确率
print('准确率:', metrics.accuracy_score(prediction,test_result))

#规范化数据，再预估准确率
z_score = StandardScaler()
train_data = z_score.fit_transform(train_data)
test_data = z_score.transform(test_data)
#用新数据做训练
new_model = svm.LinearSVC()
new_model.fit(train_data,train_result)
#重新预测
new_prediction = new_model.predict(test_data)
#准确率
print('准确率:',metrics.accuracy_score(new_prediction,test_result))

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn import svm
from sklearn import metrics

# 加载数据集，你需要把数据放到目录中
data = pd.read_csv("./breast_cancer_data-master/data.csv")

# 数据探索
# 因为数据集中列比较多，我们需要把 dataframe 中的列全部显示出来
pd.set_option('display.max_columns', None)
print(data.columns)
print(data.head(5))
print(data.describe())

# 将特征字段分成 3 组
features_mean= list(data.columns[2:12])
features_se= list(data.columns[12:22])
features_worst=list(data.columns[22:32])

# 数据清洗
# ID 列没有用，删除该列
data.drop("id",axis=1,inplace=True)
# 将 B 良性替换为 0，M 恶性替换为 1
data['diagnosis']=data['diagnosis'].map({'M':1,'B':0})

# 将肿瘤诊断结果可视化
sns.countplot(data['diagnosis'],label="Count")
plt.show()
# 用热力图呈现 features_mean 字段之间的相关性
corr = data[features_mean].corr()
plt.figure(figsize=(14,14))
# annot=True 显示每个方格的数据
sns.heatmap(corr, annot=True)
plt.show()

# 特征选择
#features_remain = ['radius_mean','texture_mean', 'smoothness_mean','compactness_mean','symmetry_mean', 'fractal_dimension_mean']
features_remain = data.columns[1:31]
print(features_remain)
print('-'*100)

# 抽取 30% 的数据作为测试集，其余作为训练集
train, test = train_test_split(data, test_size = 0.3)# in this our main data is splitted into train and test
# 抽取特征选择的数值作为训练和测试数据
train_X = train[features_remain]
train_y=train['diagnosis']
test_X= test[features_remain]
test_y =test['diagnosis']

# 采用 Z-Score 规范化数据，保证每个特征维度的数据均值为 0，方差为 1
ss = StandardScaler()
train_X = ss.fit_transform(train_X)
test_X = ss.transform(test_X)


# 创建 SVM 分类器
model = svm.LinearSVC()
# 用训练集做训练
model.fit(train_X,train_y)
# 用测试集做预测
prediction=model.predict(test_X)
print('准确率: ', metrics.accuracy_score(prediction,test_y))

------
准确率: 0.9649122807017544

# -*- coding: utf-8 -*-
"""
Created on Sun Mar 17 23:18:31 2019

@author: xcma1
"""
 
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn import metrics

# 加载数据集，你需要把数据放到目录中
data = pd.read_csv("./data.csv")
# 数据探索
# 因为数据集中列比较多，我们需要把 dataframe 中的列全部显示出来
pd.set_option('display.max_columns', None)
# 将特征字段分成 3 组
features_mean= list(data.columns[2:12])
features_se= list(data.columns[12:22])
features_worst=list(data.columns[22:32])
# 数据清洗

# ID 列没有用，删除该列
data.drop("id",axis=1,inplace=True)

# 将 B 良性替换为 0，M 恶性替换为 1
data['diagnosis']=data['diagnosis'].map({'M':1,'B':0})

# 特征选择
features_remain = ['radius_mean', 'texture_mean', 'perimeter_mean',
       'area_mean', 'smoothness_mean', 'compactness_mean', 'concavity_mean',
       'concave points_mean', 'symmetry_mean', 'fractal_dimension_mean',
       'radius_se', 'texture_se', 'perimeter_se', 'area_se', 'smoothness_se',
       'compactness_se', 'concavity_se', 'concave points_se', 'symmetry_se',
       'fractal_dimension_se', 'radius_worst', 'texture_worst',
       'perimeter_worst', 'area_worst', 'smoothness_worst',
       'compactness_worst', 'concavity_worst', 'concave points_worst',
       'symmetry_worst', 'fractal_dimension_worst']

# 抽取 30% 的数据作为测试集，其余作为训练集
train, test = train_test_split(data, test_size = 0.3)# in this our main data is splitted into train and test

# 抽取特征选择的数值作为训练和测试数据
train_X = train[features_remain]
train_y=train['diagnosis']
test_X= test[features_remain]
test_y =test['diagnosis']

# 采用 Z-Score 规范化数据，保证每个特征维度的数据均值为 0，方差为 1
ss = StandardScaler()
train_X = ss.fit_transform(train_X)
test_X = ss.transform(test_X)

# 创建 SVM 分类器
model = svm.SVC()
# 用训练集做训练
model.fit(train_X,train_y)
# 用测试集做预测
prediction=model.predict(test_X)
print('准确率: ', metrics.accuracy_score(prediction,test_y))

通过多次运行，发现每次结果也不太一样，但是大体上，svc的效果好于linearsvc，多特征好于少特征。

附代码如下：
#-coding=utf-8
from sklearn import svm
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn import metrics

def train_func(func, features_remain, desc):
    #分类阶段：模型训练
    #抽取30%的数据作为测试集，其余作为训练集
    train, test = train_test_split(data, test_size=0.3)
    train_x = train[features_remain]
    train_y = train['diagnosis']
    test_x = test[features_remain]
    test_y = test['diagnosis']

    ss = StandardScaler()
    train_x = ss.fit_transform(train_x)
    test_x = ss.transform(test_x)

    #创建SVM分类器
    if(func == "linear"):
        model = svm.LinearSVC()
    else:
        model = svm.SVC()
    #用训练集做训练
    model.fit(train_x, train_y)
    #用测试集做预测

    #分类阶段：模型评估
    predict_y = model.predict(test_x)
    infos1 = "测试集准确率：" + str(metrics.accuracy_score(predict_y, test_y))

    predict_yy = model.predict(train_x)
    infos2 = "训练集准确率：" + str(metrics.accuracy_score(predict_yy, train_y))
    print desc + ":" + infos1 + " " + infos2


#准备阶段：数据探索
data = pd.read_csv('./breast_cancer_data-master/data.csv')
#把所有的列都显示出来（在打印的时候）
pd.set_option('display.max_columns', None)

features_mean = list(data.columns[2:12])
features_se = list(data.columns[12:22])
features_worst = list(data.columns[22:32])

#准备阶段：数据清洗，id列没有用，删除该列
data.drop("id", axis=1, inplace=True)

#准备阶段：数据可视化
sns.countplot(data['diagnosis'], label='Count')
plt.show()
corr = data[features_mean].corr()
plt.figure(figsize=(14,14))
sns.heatmap(corr, annot=True)
plt.show()

#分类阶段：特征选择
features_remain = ['radius_mean', 'texture_mean', 'smoothness_mean', 'compactness_mean', 'symmetry_mean', 'fractal_dimension_mean']

#分类阶段：模型训练+模型评估
train_func('svc', features_remain, 'svc_six')
train_func('svc', features_mean, 'svc_all')
train_func('linear', features_remain, 'linearsvc_six')
train_func('linear', features_mean, 'linearsvc_all')