基于逻辑回归识别坐标是否在第一象限

这个难到是不难，主要是熟悉python和文件IO

第一次写python，模块化做的不是很好，代码写的也比较乱。

也没有使用numpy，这代码就算python的第一次练手了

#!/usr/bin/python3.5
import os
import math

def Read_W_and_b(m):
    if not os.path.exists('W_and_b.txt'):
        os.mknod('W_and_b.txt')
        f_W_and_b = open('W_and_b.txt','a')

        W_and_b = []
        for i in range(m):
            f_W_and_b.write('0.0')
            f_W_and_b.write(' ')
            W_and_b.append(0)

        W_and_b.append(0)

        f_W_and_b.write('0')
        f_W_and_b.write('
')
        f_W_and_b.close()
        return W_and_b
    else:
        f_W_and_b = open('W_and_b.txt')
        line = f_W_and_b.readline()
        result = line.split()
        W_and_b = []

        for i in result:
            W_and_b.append(float(str(i)))

        f_W_and_b.close()
        return W_and_b

def getTrainSet():
    if not os.path.exists('train.txt'):
        return -1,-1
    fTrainSet = open('train.txt')

    example = [[],[]]
    ans = []

    while 1:
        line = fTrainSet.readline()
        if not line:
            break
        pass
        result = line.split()
        example[0].append(float(str(result[0])))
        example[1].append(float(str(result[1])))
        ans.append(float(str(result[2])))
    return example,ans


def sigmoid(z):
    if z<=-100:
        return 1
    return 1/(1+math.exp(-z))


def L(a, y):
    return -(y*math.log(a)+(1-y)*math.log(1-a))

def Z(x1, x2, w1, w2,b):
    return x1*w1+x2*w2+b


def cal_d(W_and_b, X1_X2, Y):
    Derivative = [0,0,0]
    Len = len(X1_X2)-1
    for k in range(0,Len):
        z = Z(X1_X2[0][k],X1_X2[1][k],W_and_b[0],W_and_b[1],W_and_b[2])
        a = sigmoid(z)
        Derivative[0] = Derivative[0] - (a-Y[k])*X1_X2[0][k]
        Derivative[1] = Derivative[1] - (a-Y[k])*X1_X2[1][k]
        Derivative[2] = Derivative[2] - (a-Y[k])

    Derivative[0] = Derivative[0]/(Len+1)
    Derivative[1] = Derivative[1]/(Len+1)
    Derivative[2] = Derivative[2]/(Len+1)
    return Derivative

def cal_add(Derivative,alaph):
    Derivative[0] = Derivative[0] * alaph
    Derivative[1] = Derivative[1] * alaph
    Derivative[2] = Derivative[2] * alaph
    return Derivative


def Write_W_and_b(W_and_b):
    f_W_and_b = open('W_and_b.txt','w')
    f_W_and_b.write(str(W_and_b[0]))
    f_W_and_b.write(' ')
    f_W_and_b.write(str(W_and_b[1]))
    f_W_and_b.write(' ')
    f_W_and_b.write(str(W_and_b[2]))
    f_W_and_b.write('
')
    f_W_and_b.close()
    return

def train(m):#m表示数据为长度为m的行向量
    W_and_b = Read_W_and_b(m)
    X1_X2, Y = getTrainSet()

    if X1_X2 == -1 and Y == -1:
        print('无法找到训练集')
        return

    Derivative = cal_d(W_and_b, X1_X2, Y)
    add = cal_add(Derivative,1)
    W_and_b[0] = W_and_b[0] + add[0]
    W_and_b[1] = W_and_b[1] + add[1]
    W_and_b[2] = W_and_b[2] + add[2]

    print(W_and_b)
    Write_W_and_b(W_and_b)
    print("训练完成")
    return

def calPrep(x,y,W_and_b):
    save = x*W_and_b[0]+y*W_and_b[1]+W_and_b[2]
    if save >= 0:
        return 1
    else:
        return 0

def test(m):
    W_and_b = Read_W_and_b(m)
    while 1:
        x, y = map(int,input("输入x,y:").strip().split())
        if x==-1 and y==-1:
            break
        else:
             print(calPrep(x,y,W_and_b))

    return



print("开始训练: 1")
op = int(input("开始使用: 2
"))

if op==1:
    train(2)
else:
    test(2)