Python 如何在scikit学习中使用多元回归进行分类

我有一个多元回归，我用它将波段绘制成散点图，如下所示。该函数获取指数、收益率和收益率列中的投资组合列表，并在其周围放置1&2个标准差带。现在，我想重新调整相同回归的用途，并返回位于不同阴影区域的行列表

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.ticker as plticker
import matplotlib.ticker as mticker
import seaborn as sns

def all_scatter_charter(df):
    plt.figure(figsize=(12,8))
    
    #create sets for the polynomial regression and the bands
    df = df.loc[:,['returns','vol']]
    df.dropna(how='any',inplace=True)
    x = df['vol']
    y = df['returns']

    x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=0.2, random_state=42)
    x_train = x_train.values.reshape(-1, 1)
    y_train = y_train.values.reshape(-1, 1)
    y_train = y_train[x_train[:,0].argsort()]
    x_train = x_train[x_train[:, 0].argsort()]
    poly = PolynomialFeatures(degree=2)
    x_poly = poly.fit_transform(x_train)
    poly_reg = LinearRegression()
    poly_reg.fit(x_poly, y_train)
    #get STD of Y axis 
    std_y_axis = y.std()
    y2 = poly_reg.predict(x_poly)+std_y_axis
    y1 = poly_reg.predict(x_poly)-std_y_axis
    y4 = poly_reg.predict(x_poly)+(std_y_axis*2)
    y3 = poly_reg.predict(x_poly)-(std_y_axis*2)
    y2 = y2.reshape(-1)
    y1 = y1.reshape(-1)
    y3 = y3.reshape(-1)
    y4 = y4.reshape(-1)
    x_train = x_train.reshape(-1)
    

    #create chart
    plt.fill_between(x_train,y1,y2,color='#3A5675',alpha=0.3)
    plt.fill_between(x_train,y3,y4,color='#3A5675',alpha=0.1)
    ax = sns.scatterplot(data=df,y=df['returns'],x=df['vol']\
                         ,s=120,marker='o',alpha=0.3)

    
n = 200
df = pd.DataFrame(dict(
    A=np.random.randint(1, 600, size=n),
    B=np.random.randint(1, 1700, size=n)
))

df.columns = ['returns','vol']

all_scatter_charter(df)

---

对于阴影以上的任何点，我想返回'> 2SD '，在顶部光阴影> 1SD”中，在中间“中间”，在下阴影中只是一个凸点…有人有线索吗？我认为关键是从算法中提取直线方程，我知道如何使用线性回归而不是多元回归。。。