python实现Lasso回归分析(特征筛选、建模预测)

1

import numpy as np

2

import warnings

3

warnings.filterwarnings(action='ignore')

4

import pandas as pd

5

import matplotlib.pyplot as plt

6

from sklearn import metrics

7

from sklearn.metrics import mean_squared_error

8

from sklearn.linear_model import Lasso,LassoCV

9

import seaborn as sns

10

#==================读取数据=================

11

class Solution():

12

def __init__(self):

13

feature = ['男', '女', '年龄', 'CCP-正常', 'CCP-异常', 'MCV-正常', 'MCV-异常',

14

'AKA-正常', 'AKA-异常','RF-正常', 'RF-异常', 'ANA-正常', 'ANA-异常',

15

'ds-DNA-正常', 'ds-DNA-异常','CRP-正常', 'CRP-异常', 'ESR-正常', 'ESR-异常',

16

'尿蛋白-正常', '尿蛋白-异常', '尿潜血-正常', '尿潜血-异常','尿红细胞-正常',

17

'尿红细胞-异常', 'WBC-正常', 'WBC-异常', 'Hb-正常', 'Hb-异常', 'PLT-正常',

18

'PLT-异常', 'ALT-正常', 'ALT-异常', 'AST-正常', 'AST-异常', 'r-GT-正常',

19

'r-GT-异常', 'TBIL-正常', 'TBIL-异常', 'ALB-正常','ALB-异常', 'GLB-正常',

20

'GLB-异常', 'A/O-正常', 'A/O-异常', 'Cr-正常', 'Cr-异常', 'BUN-正常',

21

'BUN-异常', 'UA-正常', 'UA-异常', 'C3-正常', 'C3-异常', 'C4-正常', 'C4-异常',

22

'IgA-正常', 'IgA-异常', 'IgG-正常','IgG-异常', 'IgE-正常', 'IgE-异常',

23

'晨僵正常', '晨僵异常', '发热正常', '发热异常', '雷诺正常', '雷诺异常',

24

'口眼干正常', '口眼干异常', '头晕正常', '头晕异常', '四肢正常', '四肢异常',

25

'胸部CT正常', '胸部CT异常', '肺结节正常', '肺结节异常', '诊断结果']

26

self.feature=feature

27

28

def Data_sort(self,file):

29

data = pd.read_excel(file)

30

data = pd.DataFrame(data)

31

random_state_value = 90 # 随机种子

32

sample_number = 82 # 欠采样数目

33

def norm_2(x):

34

return (x - stats['min']) / (stats['max']-stats['min'])

35

gy_list=['年龄']

36

data_gy=data[gy_list]

37

stats = data_gy.describe()

38

stats = stats.transpose()

39

data[gy_list]=norm_2(data_gy)

40

data1 = data[self.feature]

41

data1 = data1.dropna() # 删除含缺失值的行

42

data1=data1[~data1['诊断结果'].isin([2])]

43

print(len(data1))

44

dataset=data1

45

train_dataset = dataset.sample(frac=0.7, random_state=random_state_value)

46

test_dataset = dataset.drop(train_dataset.index)

47

print(len(test_dataset))

48

train_dataset[train_dataset['诊断结果'].isin([1])]=\

49

train_dataset[train_dataset['诊断结果'].isin([1])].iloc[:sample_number]

50

train_NRA=train_dataset[train_dataset['诊断结果'].isin([0])]

51

train_RA=train_dataset[train_dataset['诊断结果'].isin([1])]

52

train_dataset=train_NRA.append(train_RA)

53

train_dataset=train_dataset.sample(frac=1,random_state=0)

54

print(len(train_dataset))

55

train_labels =train_dataset.pop('诊断结果')

56

test_labels =test_dataset.pop('诊断结果')

57

return train_dataset,train_labels,test_dataset,test_labels

58

59

#==============Lasso变量筛选====================

60

def optimal_lambda_value(self):

61

Lambdas = np.logspace(-5, 2, 200) #10的-5到10的2次方

62

# 构造空列表，用于存储模型的偏回归系数

63

lasso_cofficients = []

64

for Lambda in Lambdas:

65

lasso = Lasso(alpha = Lambda, normalize=True, max_iter=10000)

66

lasso.fit(train_dataset, train_labels)

67

lasso_cofficients.append(lasso.coef_)

68

# 绘制Lambda与回归系数的关系

69

plt.plot(Lambdas, lasso_cofficients)

70

# 对x轴作对数变换

71

plt.xscale('log')

72

# 设置折线图x轴和y轴标签

73

plt.xlabel('Lambda')

74

plt.ylabel('Cofficients')

75

# 显示图形

76

plt.show()

77

# LASSO回归模型的交叉验证

78

lasso_cv = LassoCV(alphas = Lambdas, normalize=True, cv = 10, max_iter=10000)

79

lasso_cv.fit(train_dataset, train_labels)

80

# 输出最佳的lambda值

81

lasso_best_alpha = lasso_cv.alpha_

82

print(lasso_best_alpha)

83

return lasso_best_alpha

84

85

# 基于最佳的lambda值建模

86

def model(self,train_dataset, train_labels,lasso_best_alpha):

87

lasso = Lasso(alpha = lasso_best_alpha, normalize=True, max_iter=10000)

88

lasso.fit(train_dataset, train_labels)

89

return lasso

90

91

def feature_importance(self,lasso):

92

# 返回LASSO回归的系数

93

dic={'特征':train_dataset.columns,'系数':lasso.coef_}

94

df=pd.DataFrame(dic)

95

df1=df[df['系数']!=0]

96

print(df1)

97

coef = pd.Series(lasso.coef_, index=train_dataset.columns)

98

imp_coef = pd.concat([coef.sort_values().head(10), coef.sort_values().tail(10)])

99

sns.set(font_scale=1.2)

100

# plt.rc('font', family='Times New Roman')

101

plt.rc('font', family='simsun')

102

imp_coef.plot(kind="barh")

103

plt.title("Lasso回归模型")

104

plt.show()

105

return df1

106

107

def prediction(self,lasso):

108

# lasso_predict = lasso.predict(test_dataset)

109

lasso_predict = np.round(lasso.predict(test_dataset))

110

print(sum(lasso_predict==test_labels))

111

print(metrics.classification_report(test_labels,lasso_predict))

112

print(metrics.confusion_matrix(test_labels, lasso_predict))

113

RMSE = np.sqrt(mean_squared_error(test_labels,lasso_predict))

114

print(RMSE)

115

return RMSE

116

117

if __name__=="__main__":

118

Object1=Solution()

119

train_dataset, train_labels, test_dataset, test_labels=\

120

Object1.Data_sort('F:\医学大数据课题\RA预测\RA预测\特征.xlsx')

121

lasso_best_alpha=Object1.optimal_lambda_value()

122

lasso=Object1.model(train_dataset, train_labels,lasso_best_alpha)

123

feature_choose=Object1.feature_importance(lasso)

124

RMSE=Object1.pR[-123]C[-1]:RCrediction(lasso)