机器学习实战一：knn算法 约会网站

1、sklearn库

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier

df = pd.read_table('D:\datingTestSet2.txt',header = None)
data=df.iloc[:,0:3]
target=df.iloc[:,3]
x_train,x_test, y_train, y_test =train_test_split(data,target,test_size=0.1, random_state=0)

knn=KNeighborsClassifier(weights='distance',n_neighbors=10)
knn.fit(x_train,y_train)
y_predict=knn.predict(x_test)
score=knn.score(x_test,y_test,sample_weight=None)
print(score)

输出0.75，继续学习。

2、自定义函数，up！

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split


df = pd.read_table('D:\datingTestSet2.txt',header = None)
data=df.iloc[:,0:3]
target=df.iloc[:,3]
x_train,x_test, y_train, y_test =train_test_split(data,target,test_size=0.1, random_state=0)



results=[]

x_test_array=np.array(x_test)
y_test_array=np.array(y_test)
x_train_array=np.array(x_train)
y_train_array=np.array(y_train)



def classify0(inX, dataSet, labels, k):
    dataSetSize = dataSet.shape[0]
    diffMat = tile(inX, (dataSetSize,1)) - dataSet
    sqDiffMat = diffMat**2
    sqDistances = sqDiffMat.sum(axis=1)
    distances = sqDistances**0.5
    sortedDistIndicies = distances.argsort()     
    classCount={}          
    for i in range(k):
        voteIlabel = labels[sortedDistIndicies[i]]
        classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1
    sortedClassCount = sorted(classCount.items(), key=operator.itemgetter(1), reverse=True)  #频数降序排列
    return sortedClassCount[0][0]

count=0 
for j in range(99):
    #print (x_test_array[j])

    results=classify0(x_test_array[j],x_train_array,y_train_array,10)
    #print(x_test_array[j])
    #print(y_test_array[j])
    if results ==y_test_array[j]:
        count=count+1
print(count/100)    

输出0.73，怎么还下降了，悲桑

3、分析原因

	玩游戏所耗时间	飞行常客里程数	冰淇淋	样本分类
1	0.953952	40920	8.326976	3
2	1.673904	14488	7.1534691	2
3	0.805124	26052	1.441871	1
4	0.428964	75136	13.147394	1

计算一下两个样本间的距离（欧式），公里数太大了，对计算结果的影响远远超过了其余两项，归一化吧那就！

4、sklearn+归一化

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.neighbors import KNeighborsClassifier

df = pd.read_table('D:\datingTestSet2.txt',header = None)
data=df.iloc[:,0:3]
target=df.iloc[:,3]
x_train,x_test, y_train, y_test =train_test_split(data,target,test_size=0.1, random_state=0)


def autoNorm(dataSet):
    minVals = dataSet.min(0)
    maxVals = dataSet.max(0)
    ranges = maxVals - minVals
    normDataSet = np.zeros(np.shape(dataSet))
    m = dataSet.shape[0]
    normDataSet = dataSet - np.tile(minVals, (m,1))
    normDataSet = normDataSet/np.tile(ranges, (m,1))   #element wise divide
    return normDataSet
results=[]

x_test_array=np.array(autoNorm(x_test))
y_test_array=np.array(y_test)
x_train_array=np.array(autoNorm(x_train))
y_train_array=np.array(y_train)




knn=KNeighborsClassifier(weights='distance',n_neighbors=10)
knn.fit(x_train_array,y_train_array)
y_predict=knn.predict(x_test_array)
score=knn.score(x_test_array,y_test_array,sample_weight=None)
print(score)

输出0.9，相较于0.75提升不小，非常满意。

5、自定义+归一化

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split


df = pd.read_table('D:\datingTestSet2.txt',header = None)
data=df.iloc[:,0:3]
target=df.iloc[:,3]
x_train,x_test, y_train, y_test =train_test_split(data,target,test_size=0.1, random_state=0)

def autoNorm(dataSet):
    minVals = dataSet.min(0)
    maxVals = dataSet.max(0)
    ranges = maxVals - minVals
    normDataSet = zeros(shape(dataSet))
    m = dataSet.shape[0]
    normDataSet = dataSet - tile(minVals, (m,1))
    normDataSet = normDataSet/tile(ranges, (m,1))   #element wise divide
    return normDataSet
results=[]

x_test_array=np.array(autoNorm(x_test))
y_test_array=np.array(y_test)
x_train_array=np.array(autoNorm(x_train))
y_train_array=np.array(y_train)



def classify0(inX, dataSet, labels, k):
    dataSetSize = dataSet.shape[0]
    diffMat = tile(inX, (dataSetSize,1)) - dataSet
    sqDiffMat = diffMat**2
    sqDistances = sqDiffMat.sum(axis=1)
    distances = sqDistances**0.5
    sortedDistIndicies = distances.argsort()     
    classCount={}          
    for i in range(k):
        voteIlabel = labels[sortedDistIndicies[i]]
        classCount[voteIlabel] = classCount.get(voteIlabel,0) + 1
    sortedClassCount = sorted(classCount.items(), key=operator.itemgetter(1), reverse=True)  #频数降序排列
    return sortedClassCount[0][0]

count=0 
for j in range(99):
    #print (x_test_array[j])

    results=classify0(x_test_array[j],x_train_array,y_train_array,10)
    #print(x_test_array[j])
    #print(y_test_array[j])
    if results ==y_test_array[j]:
        count=count+1
print(count/100)    
    #if results!=y_train_array
#df=pd.DataFrame({'predicted':results,'value':y_test})
#count=0
#for k in range(99):
    #if df.iloc[k,0]!=df.iloc[k,1]:
        #count=count=1

#error=count/100

输出结果0.89，just so so，复制半天不如sklearn，一把辛酸泪。

6、《机器学习实战》中分类器的错误率只有2.4%（好羡慕），莫非是k值的选取？改成3试一下。sklearn方法没有变化，依旧是0.9，自定义竟然下降到了0.88。不得不说这真是一门神奇的学科，至于10%和2.4%之间的差距，小白还要慢慢探索（能得到0.9非常开心），希望大神们指引一下~非常感谢。