scikit-learn用法总结

- “鸢尾花”的3D散点图

from sklearn import datasets
import pandas as pd 
import pyecharts as pec 
import numpy as np
#获取鸢尾花数据集
iris=datasets.load_iris()
#花瓣的长宽，茎的长宽
# print(iris.data) #150*4
# print(iris.data.shape)
# print(len(iris.data))
#花的3种类别标签
#print(iris.target)

#三类鸢尾花的3D散点图
iris_df=pd.DataFrame(
    {
    'x0':[iris.data[i][0] for i in range(len(iris.data))],
    'x1':[iris.data[i][1] for i in range(len(iris.data))],
    'x2':[iris.data[i][2] for i in range(len(iris.data))],
    'x3':[iris.data[i][3] for i in range(len(iris.data))],
    'type':iris.target
    })
df_type0=iris_df[iris_df.type==0]
# print(df_type0)
df_type1=iris_df[iris_df.type==1]
df_type2=iris_df[iris_df.type==2]

data_type0=df_type0.drop(labels=['type','x3'],axis=1).values
print(data_type0)
data_type1=df_type1.drop(labels=['type','x3'],axis=1).values
data_type2=df_type2.drop(labels=['type','x3'],axis=1).values

range_color=['#313695','#ffffbf','#a50026']
scatter3D=pec.Scatter3D('3D散点图',width=1280,height=720)
scatter3D.add('',data_type0)
scatter3D.add('',data_type1)
scatter3D.add('',data_type2)
scatter3D.render(path='iris_sdscatter.html')

- LinearSVC分类器的训练及预测

from sklearn import datasets
from sklearn import svm
from sklearn.model_selection import train_test_split
#获取鸢尾花数据集
iris=datasets.load_iris()
X=iris.data
y=iris.target
#test_size=0.3测试集占总样本个数的30%
#random_state=0确保每次都按照一样的顺序挑选数据
X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.3,random_state=0)
#LinearSVC是SVM的一种

#创建分类器/回归器
#clf.fit()
#clf.predict()

clf=svm.LinearSVC()
clf.fit(X_train,y_train)
print('测试OA：{}'.format(clf.score(X_test,y_test)))

#测试集类别标签
print(clf.predict(X_test))
print(y_test)

- “波士顿”的散点图

from sklearn import datasets
import pyecharts as pec
#波士顿数据，506行数据，13个属性+房价
boston=datasets.load_boston()
# print(boston.feature_names)
# print(boston.target.shape)
#绘制每一个‘属性-房价’的散点图
scatter_list=[]
for i in range(13):
    scatter=pec.Scatter(boston.feature_names[i],title_pos='center')
    scatter.add('',boston.data[:,i],boston.target)
    scatter_list.append(scatter)
page=pec.Page()
for i in range(13):
    page.add(scatter_list[i])
page.render()

- Lasso回归与网格搜索

from sklearn import datasets
from sklearn import linear_model
from sklearn import model_selection
from sklearn.metrics import mean_squared_error
import numpy as np

boston=datasets.load_boston()
X_train,X_test,y_train,y_test=model_selection.train_test_split(boston.data,boston.target,test_size=0.3,random_state=13)

#GridSearchCV网格搜索
alpha_range=np.arange(0,1,0.05)
param_grid={'alpha':alpha_range}
# print(param_grid)
lasso=linear_model.Lasso()
lasso_search=model_selection.GridSearchCV(lasso,param_grid)
lasso_search.fit(X_train,y_train)
# for result in lasso_search.cv_results_:#grid_scores_弃用了
#     print(result)
print(lasso_search.best_score_)
print(lasso_search.best_params_)
print(lasso_search.best_estimator_)#保存最佳的模型，以便下回调用

# lasso.fit(X_train,y_train)
# # print(lasso.predict(X_test))
# #均方根误差
# mse=mean_squared_error(y_test,lasso.predict(X_test))
# print('MSE:',mse)
# #OA
# score=lasso.score(X_test,y_test)
# print('score:',score)

'''
各种回归模型
lasso=linear_model.Lasso()
eln_reg=linear_model.ElasticNet()
svr_reg=scm.SVR()
nusvr_reg=svm.NuSVR()
ridge_reg=linear_model.Ridge()
gbr_reg=ensemble.GradientBoostingRegressor()
'''

- 特征重要性（gradientboostingregressor）之条形图

from sklearn import datasets
from sklearn import linear_model
from sklearn import model_selection
from sklearn.metrics import mean_squared_error
import numpy as np
from sklearn import ensemble
import pandas as pd 
import pyecharts as pec 
boston=datasets.load_boston()
X_train,X_test,y_train,y_test=model_selection.train_test_split(boston.data,boston.target,test_size=0.3,random_state=13)

gbr_reg=ensemble.GradientBoostingRegressor()
gbr_reg.fit(X_train,y_train)
print(gbr_reg.feature_importances_)

feature_importance_df=pd.DataFrame({
    'name':boston.feature_names,
    'importance':gbr_reg.feature_importances_
    })
feature_importance_df.sort_values(by='importance',inplace=True)
bar=pec.Bar('特征重要性',width=1280,height=720)
bar.add('',feature_importance_df['name'],feature_importance_df['importance'])
bar.render('feature_importance.html')

- 网格搜索（GridSearchCV) VS 随机搜索(RandomizedSearchCV）

from sklearn import datasets
from sklearn import linear_model
from sklearn import model_selection
from sklearn.metrics import mean_squared_error
import numpy as np
from sklearn import ensemble
import pandas as pd 
import pyecharts as pec 

boston=datasets.load_boston()
X_train,X_test,y_train,y_test=model_selection.train_test_split(boston.data,boston.target,test_size=0.3,random_state=13)
# 随机搜索
n_estimators_range=np.arange(10,1000,10)
max_depth_range=np.arange(1,10)
learning_rate_range=[0.001,0.01,0.1]
param_grid={
    'n_estimators':n_estimators_range,
    'max_depth':max_depth_range,
    'learning_rate':learning_rate_range
}
gbr_reg=ensemble.GradientBoostingRegressor()
# n_iter:测试30次
gbr_search=model_selection.RandomizedSearchCV(gbr_reg,param_grid,n_iter=30)
gbr_search.fit(X_train,y_train)
print(gbr_search.best_score_)
print(gbr_search.best_params_)
print(gbr_search.best_estimator_)

- 作业：“糖尿病”数据预测

from sklearn import datasets
import pandas as pd 
import pyecharts as pec 
from sklearn import ensemble,model_selection
import pandas as pd 
import numpy as np
from sklearn.svm import SVR
diabetes=datasets.load_diabetes()
X_train,X_test,y_train,y_test=model_selection.train_test_split(diabetes.data,diabetes.target,test_size=0.3)
# 绘制“属性-患病预测值”散点图

#print(diabetes.DESCR)#442*11,10 attributes,1 target
# diabetes_df=diabetes.data.T[:9]
# temp_list=[]
# for i in range(len(diabetes_df)):
#     temp=diabetes_df[i]
#     temp_list.append(temp)
# scatter_list=[]
# for i in range(len(diabetes_df)): 
#     scatter=pec.Scatter(diabetes.feature_names[i],title_pos='center')
#     scatter.add('',temp_list[i],diabetes.target)
#     scatter_list.append(scatter)
# page=pec.Page()
# for i in range(len(diabetes_df)):
#     page.add(scatter_list[i])
# page.render('diabetes_scatter.html')

# 首先创建回归模型，然后随机搜索参数,最后进行预测
# n_estimators_range=(10,1000,10)
# max_depth_range=np.arange(1,10)
# learning_rate_range=[0.001,0.01,0.1]
# param_grid={
#     'n_estimators':n_estimators_range,
#     'max_depth':max_depth_range,
#     'learning_rate':learning_rate_range
# }

# gbr=ensemble.GradientBoostingRegressor()
# gbr_search=model_selection.RandomizedSearchCV(gbr,param_grid,n_iter=40)
# gbr_search.fit(X_train,y_train)
# print(gbr_search.best_score_)
# print(gbr_search.best_params_)
# print(gbr_search.best_estimator_)

gamma_range=[0.05,0.1,0.2,0.3,0.4,0.5]
C_range=[1e1,1e2,1e3,1e4,1e5,1e6]
param_grid={
    'gamma':gamma_range,
    'C':C_range
}
svr=SVR(kernel='rbf')
svr_search=model_selection.RandomizedSearchCV(svr,param_grid,n_iter=40)
svr_search.fit(X_train,y_train)
print(svr_search.best_score_)
print(svr_search.best_params_)

过年每天都要运动，学习，少吃饭——2020年1月20日12:09:11