数据来源：http://archive.ics.uci.edu/ml/datasets/Wine
参考文献：《机器学习Python实战》魏贞原
博文目的：复习
工具：Geany
#导入类库
from pandas import read_csv #读数据
from pandas.plotting import scatter_matrix #画散点图
from pandas import set_option #设置打印数据精确度
import numpy as np
import matplotlib.pyplot as plt #画图
from sklearn.preprocessing import Normalizer #数据预处理：归一化
from sklearn.preprocessing import StandardScaler #数据预处理：正态化
from sklearn.preprocessing import MinMaxScaler #数据预处理：调整数据尺度
from sklearn.model_selection import train_test_split #分离数据集
from sklearn.model_selection import cross_val_score #计算算法准确度
from sklearn.model_selection import KFold #交叉验证
from sklearn.model_selection import GridSearchCV #机器学习算法的参数优化方法：网格优化法
from sklearn.linear_model import LinearRegression #线性回归
from sklearn.linear_model import Lasso #套索回归
from sklearn.linear_model import ElasticNet #弹性网络回归
from sklearn.linear_model import LogisticRegression #逻辑回归算法
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis #线性判别分析
from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis #二次判别分析
from sklearn.tree import DecisionTreeRegressor #决策树回归
from sklearn.tree import DecisionTreeClassifier #决策树分类
from sklearn.neighbors import KNeighborsRegressor #KNN回归
from sklearn.neighbors import KNeighborsClassifier #KNN分类
from sklearn.naive_bayes import GaussianNB #贝叶斯分类器
from sklearn.svm import SVR #支持向量机 回归
from sklearn.svm import SVC #支持向量机 分类
from sklearn.pipeline import Pipeline #pipeline能够将从数据转换到评估模型的整个机器学习流程进行自动化处理
from sklearn.ensemble import RandomForestRegressor #随即森林回归
from sklearn.ensemble import RandomForestClassifier #随即森林分类
from sklearn.ensemble import GradientBoostingRegressor #随即梯度上升回归
from sklearn.ensemble import GradientBoostingClassifier #随机梯度上分类
from sklearn.ensemble import ExtraTreesRegressor #极端树回归
from sklearn.ensemble import ExtraTreesClassifier #极端树分类
from sklearn.ensemble import AdaBoostRegressor #AdaBoost回归
from sklearn.ensemble import AdaBoostClassifier #AdaBoost分类
from sklearn.metrics import mean_squared_error #
from sklearn.metrics import accuracy_score #分类准确率
from sklearn.metrics import confusion_matrix #混淆矩阵
from sklearn.metrics import classification_report #分类报告
#导入数据
filename = 'wine.csv'
data = read_csv(filename, header=None, delimiter=',')
#数据理解
print(data.shape)
#print(data.dtypes)
#print(data.corr(method='pearson'))
#print(data.describe())
#print(data.groupby(0).size())
#数据可视化:直方图、散点图、密度图、关系矩阵图
#直方图
#data.hist()
#plt.show()
#密度图
#data.plot(kind='density', subplots=True, layout=(4,4), sharex=False, sharey=False)
#plt.show()
#散点图
#scatter_matrix(data)
#plt.show()
#关系矩阵图
#fig = plt.figure()
#ax = fig.add_subplot(111)
#cax = ax.matshow(data.corr(), vmin=-1, vmax=1)
#fig.colorbar(cax)
#plt.show()
#数据处理：调整数据尺度、归一化、正态化、二值化
array = data.values
X = array[:, 1:14].astype(float)
Y = array[:,0]
scaler = MinMaxScaler(feature_range=(0,1)).fit(X)
X_m = scaler.transform(X)
scaler = Normalizer().fit(X)
X_n = scaler.transform(X)
scaler = StandardScaler().fit(X)
X_s = scaler.transform(X)
#分离数据集
validation_size = 0.2
seed = 7
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=validation_size, random_state=seed)
X_m_train, X_m_test, Y_m_train, Y_m_test = train_test_split(X, Y, test_size=validation_size, random_state=seed)
X_n_train, X_n_test, Y_n_train, Y_n_test = train_test_split(X, Y, test_size=validation_size, random_state=seed)
X_s_train, X_s_test, Y_s_train, Y_s_test = train_test_split(X, Y, test_size=validation_size, random_state=seed)
#选择模型：(本例是一个分类问题)
#非线性：KNN, SVC, CART, GaussianNB,
#线性：KNN, SVR, LR, Lasso, ElasticNet, LDA,
models = {}
models['KNN'] = KNeighborsClassifier()
models['SVM'] = SVC()
models['CART'] = DecisionTreeClassifier()
models['GN'] = GaussianNB()
#models['LR'] = LinearRegression()
#models['Lasso'] = Lasso()
#models['EN'] = ElasticNet()
models['LDA'] = LinearDiscriminantAnalysis()
models['QDA'] = QuadraticDiscriminantAnalysis()
#评估模型
scoring = 'accuracy'
num_folds = 10
seed = 7
results = []
for key in models:
kfold = KFold(n_splits=num_folds, random_state=seed)
cv_results =cross_val_score(models[key], X_train, Y_train, scoring=scoring, cv=kfold)
results.append(cv_results)
print('%s %f(%f)'%(key, cv_results.mean(), cv_results.std()))
results_m = []
for key in models:
kfold = KFold(n_splits=num_folds, random_state=seed)
cv_results_m =cross_val_score(models[key], X_m_train, Y_m_train, scoring=scoring, cv=kfold)
results_m.append(cv_results_m)
print('调整数据尺度：%s %f(%f)'%(key, cv_results_m.mean(), cv_results_m.std()))
results_n = []
for key in models:
kfold = KFold(n_splits=num_folds, random_state=seed)
cv_results_n =cross_val_score(models[key], X_n_train, Y_n_train, scoring=scoring, cv=kfold)
results_n.append(cv_results_n)
print('归一化数据：%s %f(%f)'%(key, cv_results_n.mean(), cv_results_n.std()))
results_s = []
for key in models:
kfold = KFold(n_splits=num_folds, random_state=seed)
cv_results_s =cross_val_score(models[key], X_s_train, Y_s_train, scoring=scoring, cv=kfold)
results_s.append(cv_results_s)
print('正态化数据：%s %f(%f)'%(key, cv_results_s.mean(), cv_results_s.std()))
#箱线图
#算法优化：LDA
param_grid = {'solver':['svd', 'lsqr', 'eigen']}
model = LinearDiscriminantAnalysis()
kfold = KFold(n_splits=num_folds, random_state=seed)
grid = GridSearchCV(estimator=model, param_grid=param_grid, scoring=scoring, cv=kfold)
grid_result = grid.fit(X=X_train, y=Y_train)
print('最优：%s 使用：%s'%(grid_result.best_score_, grid_result.best_params_))
cv_results = zip(grid_result.cv_results_['mean_test_score'], grid_result.cv_results_['std_test_score'], grid_result.cv_results_['params'])
for mean, std, params in cv_results:
print('%f(%f) with %r'%(mean, std, params))
#算法集成
#bagging: 随机森林，极限树；
#boosting：ada， 随机梯度上升
ensembles = {}
ensembles['RF'] = RandomForestClassifier()
ensembles['ET'] = ExtraTreesClassifier()
ensembles['ADA'] = AdaBoostClassifier()
ensembles['GBM'] = GradientBoostingClassifier()
results = []
for key in ensembles:
kfold = KFold(n_splits=num_folds, random_state=seed)
cv_results =cross_val_score(ensembles[key], X_train, Y_train, scoring=scoring, cv=kfold)
results.append(cv_results)
print('%s %f(%f)'%(key, cv_results.mean(), cv_results.std()))
#集成算法调参gbm
param_grid = {'n_estimators':[10,50,100,200,300,400,500,600,700,800,900]}
model = GradientBoostingClassifier()
kfold = KFold(n_splits=num_folds, random_state=seed)
grid = GridSearchCV(estimator=model, param_grid=param_grid, cv=kfold, scoring=scoring)
grid_result = grid.fit(X=X_train, y=Y_train)
print('最优：%s 使用：%s'%(grid_result.best_score_, grid_result.best_params_))
cv_results = zip(grid_result.cv_results_['mean_test_score'], grid_result.cv_results_['std_test_score'], grid_result.cv_results_['params'])
for mean, std, params in cv_results:
print('%f(%f) with %r'%(mean, std, params))
#训练最终模型
model = LinearDiscriminantAnalysis(solver='svd')
model.fit(X=X_train, y=Y_train)
#评估最终模型
predictions = model.predict(X_test)
print(accuracy_score(Y_test, predictions))
print(confusion_matrix(Y_test, predictions))
print(classification_report(Y_test, predictions))
原文地址：http://blog.51cto.com/13542337/2057902