案例一、估计房屋价格

import numpy as np
from sklearn.tree import DecisionTreeRegressor
from sklearn.ensemble import AdaBoostRegressor
from sklearn import datasets
from sklearn.metrics import mean_squared_error, explained_variance_score   #导入评估方法
from sklearn.utils import shuffle   #打乱顺序
import matplotlib.pyplot as plt

#加载数据
house_data = datasets.load_boston()
# print(house_data.feature_names)

#打乱数据，random_state来控制打乱的顺序
x, y = shuffle(house_data.data, house_data.target, random_state=7)  

#划分训练集和测试集
num_training = int(0.8*len(x))
x_train, y_train = x[:num_training], y[:num_training]
x_test, y_test = x[num_training:], y[num_training:]

#采用决策树回归模型
dt_regression = DecisionTreeRegressor(max_depth=4)
dt_regression.fit(x_train, y_train)

#采用adaboost回归模型
ab_regression = AdaBoostRegressor(DecisionTreeRegressor(max_depth=4), n_estimators=400, random_state=7)
ab_regression.fit(x_train, y_train)

#决策树训练效果
y_pre_dt = dt_regression.predict(x_test)
mse = mean_squared_error(y_test, y_pre_dt)     #均方误差，公式是（预测-测试）平方再开方
evs = explained_variance_score(y_test, y_pre_dt)  #解释方差得分
print("\n###Decision tree performance###")
print("Mean squared error = ", round(mse, 2))
print("Explain variance score = ", round(evs, 2))

#adaBoost训练效果
y_pre_ab = ab_regression.predict(x_test)
mse = mean_squared_error(y_test, y_pre_ab)
evs = explained_variance_score(y_test, y_pre_ab)
print("##AbBoost tree performance##")
print("Mean squared error = ", round(mse, 2))  
print("Explain variance score = ", round(evs, 2))


def plot_feature_importances(feature_importances, title, feature_names):
    '''
    #画出特征的相对重要性
    :param feature_importances:
    :param title:
    :param feature_names:
    :return:
    '''
    #将重要性值标准化
    feature_importances = 100.0* (feature_importances / max(feature_importances))    
    
    #将得分从高到低排序
    index_sorted = np.flipud(np.argsort(feature_importances))
    
    #让x坐标轴上的标签居中
    pos = np.arange(index_sorted.shape[0])+0.5

    #画条形图
    plt.figure()
    plt.bar(pos, feature_importances[index_sorted], align='center')
    plt.xticks(pos, feature_names[index_sorted])
    plt.ylabel('Relative Importance')
    plt.title(title)
    plt.show()


plot_feature_importances(dt_regression.feature_importances_, 'Decission Tree  Regressor', house_data.feature_names)

plot_feature_importances(ab_regression.feature_importances_, 'AdaBoost regression', house_data.feature_names)

结果：

###Decision tree performance###
Mean squared error =  14.79
Explain variance score =  0.82
##AbBoost tree performance##
Mean squared error =  7.64

Explain variance score =  0.91

图形：

解释：采用AdaBoost算法和决策树做对比，可以看出用AdaBoost结果更好。回归决策树的最重要特征是RM，而AdaBoost最重要特征是LSTAT

涉及内容：

（1）AdaBoost算法：

点击打开链接

（2）metric

点击打开链接

（3）round（原数字，位数）函数作用是保留原数字小数点后几位

（4）feature_importances_属性是得出每个特征的占比

print(dt_regression.feature_importances_)

得到：[ 0.03856422  0.          0.          0.          0.02908806  0.62280298
  0.          0.10473383  0.          0.          0.00460542  0.

  0.20020548]

 下面代码：目的是让每个特征在100范围内展示

feature_importances = 100.0* (feature_importances / max(feature_importances))   

（5）

 index_sorted = np.flipud(np.argsort(feature_importances))

flipud():作用是让矩阵上下翻转

import numpy as np

a = [
    [1,2,3],
    [4,5,6],
    [7,8,9]
]
m = np.array(a)
#翻转90度，k是翻转的次数
mt = np.rot90(m, k=1)
print(mt)
#左右翻转
mr = np.fliplr(m)
print(mr)
#上下翻转
md = np.flipud(m)
print(md)

argsort():作用是从小到大排列