亲宝软件园·资讯

展开

python实现knn算法 使用python实现knn算法

chenge_j 人气:0
想了解使用python实现knn算法的相关内容吗,chenge_j在本文为您仔细讲解python实现knn算法的相关知识和一些Code实例,欢迎阅读和指正,我们先划重点:python,knn,下面大家一起来学习吧。

knn算法描述

对需要分类的点依次执行以下操作:
1.计算已知类别数据集中每个点与该点之间的距离
2.按照距离递增顺序排序
3.选取与该点距离最近的k个点
4.确定前k个点所在类别出现的频率
5.返回前k个点出现频率最高的类别作为该点的预测分类

knn算法实现

数据处理

#从文件中读取数据,返回的数据和分类均为二维数组
def loadDataSet(filename):
  dataSet = []
  labels = []
  fr = open(filename)
  for line in fr.readlines():
    lineArr = line.strip().split(",")
    dataSet.append([float(lineArr[0]),float(lineArr[1])])
    labels.append([float(lineArr[2])])
  return dataSet , labels


knn算法

#计算两个向量之间的欧氏距离
def calDist(X1 , X2):
  sum = 0
  for x1 , x2 in zip(X1 , X2):
    sum += (x1 - x2) ** 2
  return sum ** 0.5

def knn(data , dataSet , labels , k):
  n = shape(dataSet)[0]
  for i in range(n):
    dist = calDist(data , dataSet[i])
    #只记录两点之间的距离和已知点的类别
    labels[i].append(dist)
  #按照距离递增排序
  labels.sort(key=lambda x:x[1])
  count = {}
  #统计每个类别出现的频率
  for i in range(k):
    key = labels[i][0]
    if count.has_key(key):
      count[key] += 1
    else : count[key] = 1
  #按频率递减排序
  sortCount = sorted(count.items(),key=lambda item:item[1],reverse=True)
  return sortCount[0][0]#返回频率最高的key,即label


结果测试

已知类别数据(来源于西瓜书+虚构)

0.697,0.460,1
0.774,0.376,1
0.720,0.330,1
0.634,0.264,1
0.608,0.318,1
0.556,0.215,1
0.403,0.237,1
0.481,0.149,1
0.437,0.211,1
0.525,0.186,1
0.666,0.091,0
0.639,0.161,0
0.657,0.198,0
0.593,0.042,0
0.719,0.103,0
0.671,0.196,0
0.703,0.121,0
0.614,0.116,0

绘图方法

def drawPoints(data , dataSet, labels):
  xcord1 = [];
  ycord1 = [];
  xcord2 = [];
  ycord2 = [];
  for i in range(shape(dataSet)[0]):
    if labels[i][0] == 0:
      xcord1.append(dataSet[i][0])
      ycord1.append(dataSet[i][1])
    if labels[i][0] == 1:
      xcord2.append(dataSet[i][0])
      ycord2.append(dataSet[i][1])
  fig = plt.figure()
  ax = fig.add_subplot(111)
  ax.scatter(xcord1, ycord1, s=30, c='blue', marker='s',label=0)
  ax.scatter(xcord2, ycord2, s=30, c='green',label=1)
  ax.scatter(data[0], data[1], s=30, c='red',label="testdata")
  plt.legend(loc='upper right')
  plt.show()


测试代码

dataSet , labels = loadDataSet('dataSet.txt')
data = [0.6767,0.2122]
drawPoints(data , dataSet, labels)
newlabels = knn(data, dataSet , labels , 5)
print newlabels

运行结果

加载全部内容

相关教程
猜你喜欢
用户评论