Nearest-Neighbor Classifier紧邻算法分类器(一)

#数据来自UCI Machine Learning知识库的Iris数据集

#紧邻算法，通过计算测试集与训练集上诉特征之间的距离，分类，《Data Analysis with open Source Tools》中没有产生上述图的程序，所以，也模拟了下上述图形，但是暂#时没有平滑效果

from numpy import * 
import matplotlib.pylab as pl
train = loadtxt("D:\\iris.trn",delimiter=',',usecols=(0,1,2,3))
trainlabel= loadtxt("D:\\iris.trn",delimiter=',',usecols=(4,),dtype=str)


test = loadtxt("D:\\iris.tst",delimiter=',',usecols=(0,1,2,3))
testlabel= loadtxt("D:\\iris.tst",delimiter=',',usecols=(4,),dtype=str)


hit,miss=0,0


for i in range(test.shape[0]):
    dist = sqrt(sum((test[i]-train)**2,axis=1))
    k = argmin(dist)
    
    if trainlabel[k]== testlabel[i]:
        flag='+'
        hit +=1
    else:
        flag='-'
        miss +=1
    print flag,"\t Predicted:",trainlabel[k],"\t True:",testlabel[i]
print
print hit ,"out of",hit + miss ,"correct-Accuracy:",hit/(hit+miss+0.0)

运行结果

+ Predicted: Iris-setosa True: Iris-setosa
+ Predicted: Iris-setosa True: Iris-setosa
+ Predicted: Iris-versicolor True: Iris-versicolor
+ Predicted: Iris-versicolor True: Iris-versicolor
+ Predicted: Iris-virginica True: Iris-virginica


5 out of 5 correct-Accuracy: 1.0