주인장 – Lecture Note

회귀분석

참조 : https://daryan.tistory.com/23 sin() 함수를 그려보고, sin()함수에 난수를 더해 임의의 표본 생성

%matplotlib inline
#문서내에 그래프 출력

import numpy as np
import pylab as pl


def f(size):
    x = np.linspace(0, 4.5, size)
    y = 2 * np.sin(x * 1.5)
    return (x,y)

def sample(size):
    x = np.linspace(0, 4.5, size)
    y = 2 * np.sin(x * 1.5) + pl.randn(x.size)
    return (x,y)

pl.clf()
f_x, f_y = f(50)
pl.plot(f_x, f_y)
x, y = sample(50)
pl.plot(x, y, 'k.')

%matplotlib inline

#문서내에 그래프 출력

import numpy as np

import pylab as pl

def f(size):

x = np.linspace(0, 4.5, size)

y = 2 * np.sin(x * 1.5)

return (x,y)

def sample(size):

x = np.linspace(0, 4.5, size)

y = 2 * np.sin(x * 1.5) + pl.randn(x.size)

return (x,y)

pl.clf()

f_x, f_y = f(50)

pl.plot(f_x, f_y)

x, y = sample(50)

pl.plot(x, y, 'k.')

sklearn 모듈을 활용하여 선형회귀 : degree가 1이면 선형, 이상이면 n차 곡선이 그려짐

from sklearn.linear_model import LinearRegression

def fit_polynomial(x, y, degree): #x, y 분포에 적합한 다향식 추출
    model = LinearRegression()
    model.fit(np.vander(x, degree + 1), y)
    return model

def apply_polynomial(model, x):  #추출된 다항식에 x를 넣어 예측값 y를 구함
    degree = model.coef_.size - 1
    y = model.predict(np.vander(x, degree + 1))
    return y

model = fit_polynomial(x, y, 4)
p_y = apply_polynomial(model, x)

pl.plot(x, y, 'k.')
pl.plot(x, p_y,'g')

from sklearn.linear_model import LinearRegression

def fit_polynomial(x, y, degree): #x, y 분포에 적합한 다향식 추출

model = LinearRegression()

model.fit(np.vander(x, degree + 1), y)

return model

def apply_polynomial(model, x): #추출된 다항식에 x를 넣어 예측값 y를 구함

degree = model.coef_.size - 1

y = model.predict(np.vander(x, degree + 1))

return y

model = fit_polynomial(x, y, 4)

p_y = apply_polynomial(model, x)

pl.plot(x, y, 'k.')

pl.plot(x, p_y,'g')

degree = 1 degree = 2 degree = 3 degree = 4 numpy.vander() 함수 사용 예

#np.vander()함수 :

x = np.array([4,2,7])
print(np.vander(x, 4))
print(np.vander(x, 6))

#np.vander()함수 :

x = np.array([4,2,7])

print(np.vander(x, 4))

print(np.vander(x, 6))