Python --Read data from a numeric data file and find the multiple regression line.
A program that reads data from a tab-delimited numeric data series file as shown below and obtains multiple regression lines. 1st line: Objective variable (y) 2nd line: Explanatory variable 1 (x1) Line 3: Explanatory variable 2 (x2)
MultipleRegressionAnalysis_Data1.txt
65.7 67.8 70.3 72.0 74.3 76.2
3.27 3.06 4.22 4.10 5.26 6.18
69.7 69.7 71.3 77.6 81.0 78.7
MultipleRegressionAnalysis_2.py
from matplotlib import pyplot as plt
import numpy as np
from mpl_toolkits.mplot3d import Axes3D
def load_data(filename):
with open(filename) as lines:
return [[float(data) for data in line.strip().split('\t')]
for line in lines]
def calc_average(data):
return sum(data) / len(data)
def calc_sum_of_square(data_list):
ave = calc_average(data_list)
return sum((d - ave)**2 for data in data_list)
def calc_sum_of_square(data_list1, data_list2):
ave1 = calc_average(data_list1)
ave2 = calc_average(data_list2)
return sum((data1 * data2 - ave1 * ave2)**2 for data1, data2 in zip(data_list1, data_list2))
if __name__ == "__main__":
data = load_data('MultipleRegressionAnalysis_Data1.txt')
s = [[calc_sum_of_square(data1, data2) for data2 in data]
for data1 in data]
det = s[1][1]*s[2][2] - s[1][2]**2
#Partial regression coefficient calculation
b1 = (s[0][1]*s[2][2] - s[0][2]*s[1][2])/det
b2 = (s[0][2]*s[1][1] - s[0][1]*s[1][2])/det
b0 = calc_average(data[0]) - b1*calc_average(data[1]) - b2*calc_average(data[2])
#Multiple regression expression display
print("Y = " + str(b1) + " x1 + " + str(b2) + " x2 + " + str(b0))
fig = plt.figure()
ax = Axes3D(fig)
#Plane display from the obtained multiple regression equation
x1 = np.arange(0, 10, 0.25)
x2 = np.arange(65, 80, 0.25)
X1, X2 = np.meshgrid(x1, x2)
Y = b1 * X1 + b2 * X2 + b0
ax.plot_wireframe(X1, X2, Y)
#Original data display
ax.scatter3D(data[1], data[2], data[0], c='red')
plt.show()
result
> python MultipleRegressionAnalysis_2.py
Y = -17.118276566578807 x1 + 1.0700625679745137 x2 + 65.58796752824318
Recommended Posts