利用python进行一元线性、多项式回归（statsmodels包）

前言

• 数据使用全国大学生数学建模2021年B题（乙醇偶合制备C4烯烃）附件二数据集
• 使用python中statsmodels包实现
• 官方文档
• 参数解读

导入包

import pandas as pd
import numpy as np
import statsmodels.api as sm
import matplotlib as mpl
import matplotlib.pyplot as plt
#中文显示问题
plt.rcParams['font.sans-serif']=['SimHei']
plt.rcParams['axes.unicode_minus'] = False

# notebook嵌入图片
%matplotlib inline
# 提高分辨率
%config InlineBackend.figure_format='retina'

# 导入数据
df = pd.read_csv("C:/Users/lenovo/Desktop/附件2.csv", encoding="gbk")
df.head()

一元线性

# 时间与乙醇转化率的一元线性回归
# 添加常数项
X = sm.add_constant(df["时间"])
etoh_reg = sm.OLS(df["乙醇转化率"], X).fit()
print(etoh_reg.summary())

                            OLS Regression Results                            
==============================================================================
Dep. Variable:                  乙醇转化率   R-squared:                       0.934
Model:                            OLS   Adj. R-squared:                  0.921
Method:                 Least Squares   F-statistic:                     70.59
Date:                Sun, 03 Jul 2022   Prob (F-statistic):           0.000391
Time:                        08:09:10   Log-Likelihood:                -11.145
No. Observations:                   7   AIC:                             26.29
Df Residuals:                       5   BIC:                             26.18
Df Model:                           1                                         
Covariance Type:            nonrobust                                         
==============================================================================
                 coef    std err          t      P>|t|      [0.025      0.975]
------------------------------------------------------------------------------
const         42.6549      1.098     38.855      0.000      39.833      45.477
时间            -0.0526      0.006     -8.402      0.000      -0.069      -0.037
==============================================================================
Omnibus:                          nan   Durbin-Watson:                   1.548
Prob(Omnibus):                    nan   Jarque-Bera (JB):                0.783
Skew:                           0.564   Prob(JB):                        0.676
Kurtosis:                       1.812   Cond. No.                         362.
==============================================================================

# 时间与C4烯烃选择性的一元线性回归
# 添加常数项
X = sm.add_constant(df["时间"])
c4_reg = sm.OLS(df["C4烯烃选择性"], X).fit()
print(c4_reg.summary())

                            OLS Regression Results                            
==============================================================================
Dep. Variable:                C4烯烃选择性   R-squared:                       0.046
Model:                            OLS   Adj. R-squared:                 -0.144
Method:                 Least Squares   F-statistic:                    0.2434
Date:                Sun, 03 Jul 2022   Prob (F-statistic):              0.643
Time:                        08:09:10   Log-Likelihood:                -10.336
No. Observations:                   7   AIC:                             24.67
Df Residuals:                       5   BIC:                             24.56
Df Model:                           1                                         
Covariance Type:            nonrobust                                         
==============================================================================
                 coef    std err          t      P>|t|      [0.025      0.975]
------------------------------------------------------------------------------
const         38.5808      0.978     39.451      0.000      36.067      41.095
时间             0.0028      0.006      0.493      0.643      -0.012       0.017
==============================================================================
Omnibus:                          nan   Durbin-Watson:                   2.156
Prob(Omnibus):                    nan   Jarque-Bera (JB):                0.773
Skew:                          -0.814   Prob(JB):                        0.679
Kurtosis:                       2.938   Cond. No.                         362.
==============================================================================

二次多项式

# 时间与乙醇转化率多项式回归
X = sm.add_constant(df["时间"])
X["时间二次项"] = df["时间"]**2
etoh_ploy = sm.OLS(df["乙醇转化率"],X).fit()
print(etoh_ploy.summary())

                            OLS Regression Results                            
==============================================================================
Dep. Variable:                  乙醇转化率   R-squared:                       0.988
Model:                            OLS   Adj. R-squared:                  0.982
Method:                 Least Squares   F-statistic:                     163.2
Date:                Sun, 03 Jul 2022   Prob (F-statistic):           0.000146
Time:                        08:09:10   Log-Likelihood:                -5.2009
No. Observations:                   7   AIC:                             16.40
Df Residuals:                       4   BIC:                             16.24
Df Model:                           2                                         
Covariance Type:            nonrobust                                         
==============================================================================
                 coef    std err          t      P>|t|      [0.025      0.975]
------------------------------------------------------------------------------
const         45.2058      0.800     56.506      0.000      42.985      47.427
时间            -0.1044      0.013     -8.279      0.001      -0.139      -0.069
时间二次项          0.0002   4.15e-05      4.226      0.013    6.02e-05       0.000
==============================================================================
Omnibus:                          nan   Durbin-Watson:                   3.379
Prob(Omnibus):                    nan   Jarque-Bera (JB):                0.323
Skew:                          -0.475   Prob(JB):                        0.851
Kurtosis:                       2.550   Cond. No.                     1.26e+05
==============================================================================

# 时间与C4烯烃选择性多项式回归
# 添加常数项
X = sm.add_constant(df["时间"])
X["时间二次项"] = df["时间"]**2
c4_ploy = sm.OLS(df["C4烯烃选择性"], X).fit()
print(c4_ploy.summary())

                            OLS Regression Results                            
==============================================================================
Dep. Variable:                C4烯烃选择性   R-squared:                       0.241
Model:                            OLS   Adj. R-squared:                 -0.139
Method:                 Least Squares   F-statistic:                    0.6335
Date:                Sun, 03 Jul 2022   Prob (F-statistic):              0.577
Time:                        08:09:18   Log-Likelihood:                -9.5392
No. Observations:                   7   AIC:                             25.08
Df Residuals:                       4   BIC:                             24.92
Df Model:                           2                                         
Covariance Type:            nonrobust                                         
==============================================================================
                 coef    std err          t      P>|t|      [0.025      0.975]
------------------------------------------------------------------------------
const         39.7151      1.487     26.712      0.000      35.587      43.843
时间            -0.0203      0.023     -0.865      0.436      -0.085       0.045
时间二次项       7.798e-05   7.71e-05      1.011      0.369      -0.000       0.000
==============================================================================
Omnibus:                          nan   Durbin-Watson:                   2.491
Prob(Omnibus):                    nan   Jarque-Bera (JB):                0.686
Skew:                          -0.610   Prob(JB):                        0.710
Kurtosis:                       2.071   Cond. No.                     1.26e+05
==============================================================================

预测与绘图

# 乙醇转化率预测图
temp = np.arange(20,300,1)
c4_x_pre = np.vstack((np.ones_like(temp), temp, temp**2)).T
plt.figure(dpi = 600,figsize = (6,4))
plt.plot(df['时间'],df['乙醇转化率'],'o')
plt.plot(temp,etoh_ploy.predict(c4_x_pre))
plt.legend(['原数据','预测线'])

# C4烯烃选择性预测图
temp = np.arange(20,300,1)
c4_x_pre = np.vstack((np.ones_like(temp), temp, temp**2)).T
plt.figure(dpi = 600,figsize = (6,4))
plt.plot(df['时间'],df['C4烯烃选择性'],'o')
plt.plot(temp,c4_ploy.predict(c4_x_pre))
plt.legend(['原数据','预测线'])
plt.savefig("2.png")