机器学习（三）----决策树

模型任务

接受波士顿地区的房价特征，一共13个，
根据特征预测房子售价

数据载入

import sklearn.datasets as sd   #标准样本数据集
housing = sd.load_boston()
#使用housing.feature_names查看房价特征
#通过housing.data查看特征矩阵，

测试集和训练集划分

#打乱原数据集，保证数据随机性#随机种子random_state = 7表示用时间做种子
x,y = su.shuffle(housing.data,housing.target,random_state = 7)

train_size = int(len(x)*0.8)

train_x,test_x,train_y,test_y =x[:train_size],\
                               x[train_size:],\
                               y[:train_size],\
                               y[train_size:]

构建一般决策树模型&训练

#构建决策树模型
model = st.DecisionTreeRegressor(max_depth=4)
model.fit(train_x,train_y)
pred_test_y = model.predict(test_x)
#计算误差
print(sm.r2_score(test_y,pred_test_y))

构建正向激励的决策树模型

model_se = se.AdaBoostRegressor(st.DecisionTreeRegressor(max_depth=4),
                                n_estimators=400,
                                random_state=7)
                                model_se.fit(train_x,train_y)
pred_test_y = model_se.predict(test_x)

#计算误差
print(sm.r2_score(test_y,pred_test_y))

for test,pred_test in zip(test_y,pred_test_y):
    print(test,'->',pred_test)

查看特征影响

#提取一般决策树的特征重要性
fi_dt = model.feature_importances_
#产生降序索引
sorted_indexs = fi_dt.argsort()[::-1]   
#提取正向决策树的特征重要性
fi_ab  = model_se.feature_importances_
#产生降序索引
sorted_indexs = fi_ab.argsort()[::-1] 
# 随机森林的特征重要性也是调用feature_importances_

绘制特征重要性排序

#可视化展示
mp.figure('Feature Importace',facecolor='lightgray')
mp.subplot(211)
mp.title('Decision tree')
mp.ylabel('Importace',fontsize = 14)
mp.tick_params(labelsize = 10)
mp.grid(axis='y',linestyle = ':')
sorted_indexs = fi_dt.argsort()[::-1]   #降序索引
pos = np.arange(sorted_indexs.size)
mp.bar(pos,fi_dt[sorted_indexs],
       facecolor= 'deepskyblue',
       edgecolor = 'steelblue')
mp.xticks(pos,feature_names[sorted_indexs],rotation =30)
mp.subplot(212)
mp.title('AdaBoost Decision Tree',fontsize = 16)
mp.xlabel('Feature',fontsize = 12)
mp.ylabel('Importance',fontsize =12)
mp.tick_params(labelsize = 10)
mp.grid(axis = 'y',linestyle = ':')
sorted_indexs = fi_ab.argsort()[::-1]
pos = np.arange(sorted_indexs.size)
mp.bar(pos,fi_ab[sorted_indexs],
       facecolor= 'lightcoral',
       edgecolor = 'indianred')
mp.xticks(pos,feature_names[sorted_indexs],
          rotation = 30)
mp.tight_layout()
mp.show()