数据挖掘作业2

最新推荐文章于 2024-09-15 22:31:42 发布

ZORROyyds

最新推荐文章于 2024-09-15 22:31:42 发布

阅读量75

点赞数

文章标签：数据挖掘 python 机器学习

本文链接：https://blog.csdn.net/ZORROyyds/article/details/129479543

版权

from numpy import genfromtxt
import pandas as pd
data=pd.read_csv("D:\\boshidun.csv")
x0=data["mianji"].values.astype(int)
x1=data["woshi"].values.astype(int)
y=data["jiage"].values.astype(int)
# print(len(x0))
# print(len(y))
# print(len(x0))
# print(len(x1))
#给θ0,θ1,θ2赋值
a0= 0
a1 =0
a2 =0
#赋予学习率，迭代次数
alpha=0.00001
epochs=2000
#损失函数
def loss_error(a0,a1,a2,x0,x1,y):
    sum = 0
    m = float(len(x0))
    for i in range(0, len(x0)):
        sum +=((a0 + a1 * x0[i]+a2 * x1[i]) - y[i])**2
    return sum /(m*2)
#增加梯度下降
def train(a0,a1,a2,x0,x1,y ,alpha, epochs):
    m = float(len(x0))
    for i in range(epochs):
    #梯度值
        a0_gradient=0
        a1_gradient=0
        a2_gradient=0
        for j in range(0, len(x0)):
            a0_gradient += ((a0+a1*x0[j]+a2*x1[j])-y[j])
            a1_gradient += x0[j] * ((a0+a1*x0[j]+a2*x1[j])-y[j])
            a2_gradient += x1[j] * ((a0+a1*x0[j]+a2*x1[j])-y[j])
        #防止出现overflow encountered in multiply
        a0 = a0-(alpha * (1/m)*a0_gradient)
        a1 = a1-(alpha * (1/m)*a1_gradient)
        a2 = a2-(alpha * (1/m)*a2_gradient)
    return a0,a1,a2
print("loss_error:",loss_error(a0,a1,a2,x0,x1,y))
a0,a1,a2= train(a0,a1,a2,x0,x1,y ,alpha, epochs)#赋予新的参数
print("赋予新的参数后得到的新的loss_error:",loss_error(a0,a1,a2,x0,x1,y))