快速入门Python机器学习（35）

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.neural_network import MLPRegressor
from sklearn.neighbors import KNeighborsRegressor
from sklearn.preprocessing import OneHotEncoder#独热编码，同pd.get_dummies，但仅能用于整型变量
def base_for_dummies():
        colour = pd.DataFrame({'数据特征':[1,2,3,4,5,6,7],'颜色类型':['赤','橙','黄','绿','青','蓝','紫']})
        print(colour)

        colour_dum = pd.get_dummies(colour)
        print(colour_dum)

  colour['数据特征'] = colour['数据特征'].astype(str)
        colour_dum = pd.get_dummies(colour,columns=['数据特征'])
        print(colour_dum)

class sklearn.preprocessing.OneHotEncoder(*, categories='auto', drop=None, sparse=True, dtype=, handle_unknown='error')

#生成随机数列
rnd = np.random.RandomState(56)
x = rnd.uniform(-5,5,size=50)
#向数据添加噪音
y_no_noise = (np.cos(6*x)+x)
X = x.reshape(-1,1)
y = (y_no_noise + rnd.normal(size=len(x)))/2
plt.plot(X,y,'o',c='g')
plt.show()

        line = np.linspace(-5,5,1000,endpoint=False).reshape(-1,1)
        mlpr = MLPRegressor().fit(X,y)
        knr = KNeighborsRegressor().fit(X,y)
        plt.plot(line,mlpr.predict(line),label='MLP')
        plt.plot(line,knr.predict(line),label='KNN')
        plt.plot(X,y,'o',c='g')
        plt.legend(loc='best')
        plt.show()

#设置11个箱子
bins = np.linspace(-5,5,11)
#将数据装箱
target_bin = np.digitize(X, bins=bins)
print("装箱数据范围：\n{}".format(bins))
print("前10个数据点的特征值：\n{}".format(X[:10]))
print("前10个数据点所在的箱子：\n{}".format(target_bin[:10]))

分为
[-5,-4],[-4,-3],[-3,-2],[-2,-1],[-1,0],[0,1],[1,2],[2,3],[3,4],[4,5] 
10个箱子
装箱数据范围：
[-5. -4. -3. -2. -1.  0.  1.  2.  3.  4.  5.]
前10个数据点的特征值：
[[ 4.84191851]
 [-1.66587734]
 [ 1.73701625]
 [-3.03609583]
 [-1.45553535]
 [ 3.13365953]
 [-2.52150446]
 [-0.41388675]
 [ 3.77301421]
 [-1.08341539]]
前10个数据点所在的箱子：
[[10]
 [ 4]
 [ 7]
 [ 2]
 [ 4]
 [ 9]
 [ 3]
 [ 5]
 [ 9]
 [ 4]]

onehot = OneHotEncoder(sparse = False)
onehot.fit(target_bin)
#将独热编码转为数据
X_in_bin = onehot.transform(target_bin)
print("装箱后的数据形态：\n{}".format(X_in_bin.shape))
print("装箱后的前10个数据点：\n{}".format(X_in_bin[:10]))

装箱后的数据形态：
(50, 10)
装箱后的前10个数据点：
[[0. 0. 0. 0. 0. 0. 0. 0. 0. 1.]
 [0. 0. 0. 1. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 1. 0. 0. 0.]
 [0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 1. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
 [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 1. 0. 0. 0. 0. 0.]
 [0. 0. 0. 0. 0. 0. 0. 0. 1. 0.]
 [0. 0. 0. 1. 0. 0. 0. 0. 0. 0.]]

#使用独热编码进行数据表达
onehot_line = onehot.transform(np.digitize(line,bins=bins))
onehot_mlpr = MLPRegressor().fit(X_in_bin,y)
onehot_knr = KNeighborsRegressor().fit(X_in_bin,y)
plt.plot(line,onehot_mlpr.predict(onehot_line),label='New MLP')
plt.plot(line,onehot_knr.predict(onehot_line),label='New KNN')
plt.plot(X,y,'o',c='g')
plt.legend(loc='best')
plt.show()

class sklearn.preprocessing.PolynomialFeatures(degree=2, *, interaction_only=False, include_bias=True, order='C')

array_1 = [0,1,2,3,4]
array_2 = [5,6,7,8,9]
array3 = np.hstack((array_1,array_2))
print("将数组2添加到数据1后面去得到:\n{}".format(array3))

将数组2添加到数据1后面去得到:
[0 1 2 3 4 5 6 7 8 9]

#将原始数据和装箱数据进行堆叠
X_stack = np.hstack([X,X_in_bin])
print("X.shape:\n",X.shape)
print("X_in_bin.shape:\n",X_in_bin.shape)
print("X_stack.shape:\n",X_stack.shape

X.shape:
 (50, 1)
X_in_bin.shape:
 (50, 10)
X_stack.shape:
 (50, 11)

#将数据进行堆叠
line_stack = np.hstack([line,onehot_line])
mlpr_interact = MLPRegressor().fit(X_stack,y)
plt.plot(line,mlpr_interact.predict(line_stack),label='MLP for interaction')
plt.ylim(-4,4)
for vline in bins:
       plt.plot([vline,vline],[-5,5],":",c='k')
plt.legend(loc='lower right')
plt.plot(X,y,"o",c='g')
plt.show()

#使用新的叠堆方式处理数据
X_multi = np.hstack([X_in_bin,X*X_in_bin])
print("X_multi.shape:\n",X_multi.shape)
print("X_multi[0]:\n",X_multi[0])

X_multi.shape:
 (50, 20)
X_multi[0]:
 [0.         0.         0.         0.         0.         0.
 0.         0.         0.         1.         0.         0.
 0.         0.         0.         0.         0.         0.
 0.         4.84191851]
前10个数据点的特征值：
[[ 4.84191851]
 [-1.66587734]
 [ 1.73701625]
 [-3.03609583]
 [-1.45553535]
 [ 3.13365953]
 [-2.52150446]
 [-0.41388675]
 [ 3.77301421]
 [-1.08341539]]

rnd = np.random.RandomState(50)
x = rnd.uniform(-5,5,size=50)
X = x.reshape(-1,1)
y_no_noise = (np.cos(6*x)+x)
y = (y_no_noise + rnd.normal(size=len(x)))/2
poly = PolynomialFeatures(degree=20,include_bias=False)
X_poly = poly.fit_transform(X)
print(X_poly.shape)

(50, 20)

print("原始数据第一个样本：\n {} ".format(X[0]))
print("多项式处理后第一个样本：\n{} ".format(X_poly[0]))
print("PolynomialFeatures对原始数据的处理：\n{}".format(poly.get_feature_names()))

原始数据第一个样本：
[4.84191851]
多项式处理后第一个样本：
[4.84191851e+00 2.34441748e+01 1.13514784e+02 5.49629333e+02
 2.66126044e+03 1.28856062e+04 6.23910549e+04 3.02092403e+05
 1.46270680e+06 7.08230711e+06 3.42919539e+07 1.66038846e+08
 8.03946561e+08 3.89264373e+09 1.88478637e+10 9.12598202e+10
 4.41872612e+11 2.13951118e+12 1.03593388e+13 5.01590740e+13]
PolynomialFeatures对原始数据的处理：
['x0', 'x0^2', 'x0^3', 'x0^4', 'x0^5', 'x0^6', 'x0^7', 'x0^8', 'x0^9', 'x0^10', 'x0^11', 'x0^12', 'x0^13', 'x0^14', 'x0^15', 'x0^16', 'x0^17', 'x0^18', 'x0^19', 'x0^20']

line = np.linspace(-5,5,1000,endpoint=False).reshape(-1,1)
LNR_poly = LinearRegression().fit(X_poly,y)
line_poly = poly.transform(line)
plt.plot(line,LNR_poly.predict(line_poly),label='Line Regressor')
plt.xlim(np.min(X)-0.5,np.max(X)+0.5)
plt.ylim(np.min(y)-0.5,np.max(y)+0.5)
plt.plot(X,y,"o",c='g')
plt.legend(loc='lower right')
plt.show()

class sklearn.feature_selection.SelectPercentile(score_func=, *, percentile=10)[source]

def dealdata(mydata):
    if str(mydata)[-1]=='%':
        try:
            return (float(str(mydata)[:-1])/100)
        except:
            return mydata 
    elif str(mydata)[-1]=='万':
        try:
            return float(str(mydata)[:-1])*10000
        except:
            return mydata
    elif str(mydata)[-1]=='亿':
        try:
            return float(str(mydata)[:-1])*100000000
        except:
            return mydata 
     else:
            return mydata      
def dealcsv():
       stock = pd.read_csv('000002.csv',encoding='GBK')
       new_stock = stock.applymap(dealdata)
       print(new_stock.head())
       new_stock.to_csv('stock.csv',encoding='GBK')
from sklearn.preprocessing import StandardScaler
from sklearn.feature_selection import SelectPercentile
def stock():
    stock = pd.read_csv('stock.csv',encoding='GBK')
    print(stock.head())

#设置目标
y = stock['涨跌幅']
print(y.shape)
print(y[0])

(408,)
-0.0252

#提取特征值
features = stock.loc[:,'价格':'流通市值']
X = features.values
print(X.shape)
print(X[1])

(408, 15)
[ 2.594e+01 -6.700e-01  0.000e+00  2.661e+01  2.661e+01  2.670e+01
  2.593e+01  6.581e+03  1.721e+07  3.380e-02  9.800e-01  3.226e-01
  2.890e-02  2.950e+01  5.020e+08]

mlp = MLPRegressor(random_state=53,hidden_layer_sizes=[100,100,100],alpha=0.1)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=62)
#预处理
scaler = StandardScaler()
scaler.fit(X_train)
X_train_scaled = scaler.transform(X_train)
X_test_scaled = scaler.transform(X_test)
mlp.fit(X_train_scaled,y_train)
print('训练集得分:{:.2%}'.format(mlp.score(X_train_scaled,y_train)))
print('测试集得分:{:.2%}'.format(mlp.score(X_test_scaled,y_test)))

训练集得分:47.47%
测试集得分:48.47%

wanted = stock.loc[:,'名称']
print(wanted[y>=-0.05])

0      当代文体
1      鼎通科技
2      澜起科技
3       苏博特
4      亿利洁能
       ... 
362    盟升电子
363    青海春天
364    瀚蓝环境
365     华润微
366    ST亚星
Name: 名称, Length: 367, dtype: object

from sklearn.preprocessing import PolynomialFeatures
select = SelectPercentile(percentile=50)
select.fit(X_train_scaled,y_train)
X_train_selected = select.transform(X_train_scaled)
print('经过缩放后的形态:{}'.format(X_train_scaled.shape))
print('特征选择后的形态:{}'.format(X_train_selected.shape))
15个特征项成为了7个，某些被过滤掉了
mask = select.get_support()
print(mask)
#用图像表示单一变量法特征选择结果
plt.matshow(mask.reshape(1,-1),cmap=plt.cm.cool)
plt.xlabel(u"特征选择")
plt.rcParams['font.sans-serif']=['SimHei']
plt.rcParams['axes.unicode_minus']=False
plt.show()

[False False False False False False False True True True True False True True True]

#使用特征选择后数据集训练神经网络
X_test_selected = select.transform(X_test_scaled)
mlp_select = MLPRegressor(random_state=53,hidden_layer_sizes=[100,100,100],alpha=0.1)
mlp_select.fit(X_train_selected,y_train)
print('单一变量法特征选择后训练集得分:{:.2%}'.format(mlp_select.score(X_train_selected,y_train)))
print('单一变量法特征选择后测试集得分:{:.2%}'.format(mlp_select.score(X_test_selected,y_test)))

单一变量法特征选择后训练集得分:29.44%
单一变量法特征选择后测试集得分:18.36%

class sklearn.feature_selection.SelectFromModel(estimator, *, threshold=None, prefit=False, norm_order=1, max_features=None, importance_getter='auto')

from sklearn.feature_selection import SelectFromModel
from sklearn.ensemble import RandomForestRegressor
Def selectFromModel():
    stock = pd.read_csv('stock.csv',encoding='GBK')
    y = stock['涨跌幅’]
    features = stock.loc[:,'价格':'流通市值’]
    X = features.values
    X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=62)
    #预处理
    scaler = StandardScaler()
    scaler.fit(X_train)
    X_train_scaled = scaler.transform(X_train)
    X_test_scaled = scaler.transform(X_test)
    sfm = SelectFromModel(estimator=RandomForestRegressor(n_estimators=100,random_state=38),threshold='median')
    sfm.fit(X_train_scaled,y_train)
    X_train_sfm = sfm.transform(X_train_scaled)                    
    print('经过随机森林模型特征选择后的的数据形态:{}'.format(X_train_sfm.shape))
    mask = sfm.get_support()
    print(mask)

经过随机森林模型进行特征后的的数据形态:(306, 8)
[False True False False False False True False False True True False True True True True]

#使用随机森林特征选择后数据集训练神经网络
X_test_sfm = sfm.transform(X_test_scaled)
mlp_sfm = MLPRegressor(random_state=53,hidden_layer_sizes=[100,100,100],alpha=0.1)
mlp_sfm.fit(X_train_sfm,y_train)
print('经过随机森林模型特征选择后训练集得分:{:.2%}'.format(mlp_sfm.score(X_train_sfm,y_train)))
print('经过随机森林模型特征选择后测试集得分:{:.2%}'.format(mlp_sfm.score(X_test_sfm,y_test)))

经过随机森林模型特征选择后的数据形态:(306, 8)
[False True False False False False False True True True True True True True False]
经过随机森林模型特征选择后训练集得分:37.46%
经过随机森林模型特征选择后测试集得分:34.72%

class sklearn.feature_selection.RFE(estimator, *, n_features_to_select=None, step=1, verbose=0, importance_getter='auto')

from sklearn.feature_selection import RFE
def elimination():
    stock = pd.read_csv('stock.csv',encoding='GBK’)
    y = stock['涨跌幅’]
    features = stock.loc[:,'价格':'流通市值’]
    X = features.values
    X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=62)
    #预处理
    scaler = StandardScaler()
    scaler.fit(X_train)
    X_train_scaled = scaler.transform(X_train)
    X_test_scaled = scaler.transform(X_test)
    rfe = RFE(RandomForestRegressor(n_estimators=100,random_state=38),n_features_to_select=8)
    rfe.fit(X_train_scaled,y_train)
    X_train_rfe = rfe.transform(X_train_scaled)
    print('经过随机森林模型进行迭代特征选择后的的数据形态:{}'.format(X_train_rfe.shape))
mask = rfe.get_support()
    print(mask)
    print(mask)
     #用图像表示特征选择结果
    plt.matshow(mask.reshape(1,-1),cmap=plt.cm.cool)
    plt.xlabel(u"特征选择")
    plt.rcParams['font.sans-serif']=['SimHei’]
    plt.rcParams['axes.unicode_minus']=False
    plt.show()

经过随机森林模型进行迭代特征选择后的数据形态:(306, 8)
[False True False True False False False True False True True True True True False]

       X_test_rfe = rfe.transform(X_test_scaled)
       mlp_rfe = MLPRegressor(random_state=53,hidden_layer_sizes=[100,100,100],alpha=0.1)
       mlp_rfe.fit(X_train_rfe,y_train)
        print('经过随机森林迭代特征选择后训练集得分:{:.2%}'.format(mlp_rfe.score(X_train_rfe,y_train)))
        print('经过随机森林迭代特征选择后测试集得分:{:.2%}'.format(mlp_rfe.score(X_test_rfe,y_test)))

随机森林迭代特征选择后训练集得分:41.59%
经过随机森林迭代特征选择后测试集得分:43.61%