Python实现孤立森林（IForest）+SVR的组合预测模型

只讨论性能，不考虑关联性，降噪数据未填补。

1.引入数据集

from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
from sklearn.metrics import classification_report
from sklearn.metrics import explained_variance_score
from sklearn.ensemble import RandomForestRegressor
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
import pandas as pd
import numpy as np
import time

from sklearn import metrics

import csv

data=[]
traffic_feature=[]
traffic_target=[]
csv_file = csv.reader(open('turang.csv'))
for content in csv_file:
    content=list(map(float,content))
    if len(content)!=0:
        data.append(content)
        traffic_feature.append(content[0:4])
        traffic_target.append(content[-1])

data=np.array(data)

traffic_feature=np.array(traffic_feature)
traffic_target=np.array(traffic_target)

df=pd.Dataframe(data=data,columns = ['土壤温度','空气湿度','空气温度','光照强度','土壤水分'])
df

3083 rows × 5 columns

2.标准化

scaler = StandardScaler() # 标准化转换
scaler.fit(traffic_feature)  # 训练标准化对象
traffic_feature= scaler.transform(traffic_feature)   # 转换数据集
feature_train,feature_test,target_train, target_test = train_test_split(traffic_feature,traffic_target,test_size=0.1,random_state=10)

3.SVR

from sklearn.svm import SVR
import matplotlib.pyplot as plt  
start1=time.time()
model_svr = SVR(C=1,epsilon=0.1,gamma=10)
model_svr.fit(feature_train,target_train)
predict_results1=model_svr.predict(feature_test)
end1=time.time()

SVRRESULT=predict_results1

plt.plot(target_test)#测试数组
plt.plot(predict_results1)#测试数组
plt.legend(['True','SVR'])
fig = plt.gcf()
fig.set_size_inches(18.5, 10.5)
plt.title("SVR")  # 标题
plt.show()
print("EVS:",explained_variance_score(target_test,predict_results1))
print("R2:",metrics.r2_score(target_test,predict_results1))
print("Time:",end1-start1)

EVS: 0.6563954626493961
R2: 0.6536622059850858
Time: 0.3603041172027588

4.引入IForest降噪

from sklearn.ensemble import IsolationForest

model_isof = IsolationForest()
outlier_label = model_isof.fit_predict(df)
# 将array 类型的标签数据转成 Dataframe
outlier_pd = pd.Dataframe(outlier_label, columns=['outlier_label'])
# 将标签数据与原来的数据合并
data_merge = pd.concat((df, outlier_pd), axis=1)
normal_source = data_merge[data_merge['outlier_label']==1]
normal_source

 5.测试效果

IF_traffic_feature=normal_source.values[:,[0,1,2,3]]
IF_traffic_target=normal_source.values[:,[4]]
scaler.fit(IF_traffic_feature)  # 训练标准化对象
IF_traffic_feature= scaler.transform(IF_traffic_feature)   # 转换数据集
feature_train,feature_test,target_train, target_test = train_test_split(IF_traffic_feature,IF_traffic_target,test_size=0.1,random_state=10)
from sklearn.svm import SVR
import matplotlib.pyplot as plt  
start1=time.time()
model_svr = SVR(C=1,epsilon=0.1,gamma=10)
model_svr.fit(feature_train,target_train)
predict_results1=model_svr.predict(feature_test)
end1=time.time()


plt.plot(target_test)#测试数组
plt.plot(predict_results1)#测试数组
plt.legend(['True','SVR'])
fig = plt.gcf()
fig.set_size_inches(18.5, 10.5)
plt.title("SVR")  # 标题
plt.show()
print("EVS:",explained_variance_score(target_test,predict_results1))
print("R2:",metrics.r2_score(target_test,predict_results1))
print("Time:",end1-start1)

EVS: 0.5487895968648998
R2: 0.5255880566754088
Time: 0.2094569206237793

效果不怎么样...可能是数据集或者未数据填补的原因。