k-近邻居算法 识别数字

k-近邻居算法 识别数字 

k-近邻算法

有已经分类好的训练集，和一些需要你去分类的数据（暂且叫做目标）。欲分类，只需要算出 目标与各个训练集的距离，在找出距离最近的 k 个训练数据，这k个数据中哪种数据多，就将目标确定为什么种类。

数字样式如下：我们有数百个这种记录数字形状的txt文档，样式分别是1~9

运用这些文档对计算机进行训练，使之以后可以自动判断此类文档表示什么数字

我的代码展示，判断准确率为90%，有待提高

import os
import numpy as np


list_path = os.listdir('D:\aaa编程之pycharm\要来了吗，机器学习！\k-近邻\testDigits')
check_path = os.listdir('D:\aaa编程之pycharm\要来了吗，机器学习！\k-近邻\测试集')
print(check_path)


def read_allfile(list_path):
    list_number = len(list_path)
    returnData = np.zeros((list_number,32*32))
    list_order = 0
    returnlabel = []  # 训练集标签
    for filepath in list_path:
        fr = open('D:\aaa编程之pycharm\要来了吗，机器学习！\k-近邻\testDigits\'+filepath)
        returnlabel.append(filepath[0])
        for i in range(32):
            line = fr.readline()
            for j in range(32):
                returnData[list_order,32*i+j] = int(line[j])
        list_order += 1
    return returnData,returnlabel
def read_allfile2(list_path):
    list_number = len(list_path)
    returnData = np.zeros((list_number,32*32))
    list_order = 0
    returnlabel = []  # 训练集标签
    for filepath in list_path:
        fr = open('D:\aaa编程之pycharm\要来了吗，机器学习！\k-近邻\测试集\'+filepath)
        returnlabel.append(filepath[0])
        for i in range(32):
            line = fr.readline()
            for j in range(32):
                returnData[list_order,32*i+j] = int(line[j])
        list_order += 1
    return returnData,returnlabel
train_Data,train_label = read_allfile(list_path)
check_Data,check_label = read_allfile2(check_path)
print(check_Data)
print(check_label)
def classify(train_data,check_data,train_labels,k):
    train_number = train_data.shape[0]
    gass_label = []
    for line in check_data:
        temp_Mat = np.tile(line,(train_number,1))
        temp_Mat = (temp_Mat-train_data)**2
        distance = temp_Mat.sum(axis = 1)
        sort_number = distance.argsort()

        dict = {}
        for j in range(k):
            dict[train_labels[sort_number[j]]] = dict.get(train_labels[sort_number[j]],0) + 1
        # print(dict)
        dict = list(dict.items())
        # print(dict)
        temp_Mat = dict[0]
        for max in dict:
            if max[1] >= temp_Mat[1]:
                temp_Mat = max
        gass_label.append(temp_Mat[0])
    return gass_label

answer = classify(train_Data,check_Data,train_label,50)
print(answer)
def check(true,gass):
    chang = len(true)
    right_number = 0
    for i in range(chang):
        if true[i] == gass[i]:
            right_number += 1
    return right_number/chang
how = check(check_label,answer)
print(how)