邻近算法,或者说K最近邻(kNN,k-NearestNeighbor)分类算法是数据挖掘分类技术中最简单的方法之一。所谓K最近邻,就是k个最近的邻居的意思,说的是每个样本都可以用它最接近的k个邻居来代表。
kNN算法的核心思想是如果一个样本在特征空间中的k个最相邻的样本中的大多数属于某一个类别,则该样本也属于这个类别,并具有这个类别上样本的特性。
数据预备,这里使用random函数生成10*2的矩阵作为两列特征值,1个10个元素数组作为类别值
import numpy as npimport matplotlib.pyplot as plt x_train = np.random.rand(10,2)*8y_train = np.random.randint(0,2,10) x = np.array([3,4]) k=3plt.scatter(x_train[y_train==1,0],x_train[y_train==1,1],color="red") plt.scatter(x_train[y_train==0,0],x_train[y_train==0,1],color="green") plt.scatter(x[0],x[1],marker='+',color="blue") plt.show()
绿点为类别0,红点为类别1
X_train = np.array(x_train) Y_train = np.array(y_train) from math import sqrtdistances = []for x_train in X_train: d = sqrt(np.sum((x-x_train)**2)) distances.append(d) distances = [sqrt(np.sum((x-x_train)**2)) for x_train in X_train] argindex = np.argsort(distances) from collections import Counter topK_Y = [Y_train[i] for i in argindex[:k]] votes = Counter(topK_Y) votes.most_common(1)[0][0]
执行结果为判断x点大概率为类别0(绿点)
使用sklearn中封装的knn算法
from sklearn.neighbors import KNeighborsClassifier knn_clf = KNeighborsClassifier(n_neighbors=3) knn_clf.fit(X_train,Y_train) knn_clf.predict(x.reshape(1,-1))[0]
封装自己的knn算法
# _*_ encoding:utf-8 _*_import numpy as np from math import sqrt from collections import Counterclass KNNClassifier: def __init__(self,k): assert k>=1, "k must be valid" self.k = k self._X_train = None self._Y_train = None def fit(self,X_train,Y_train): assert X_train.shape[0] == Y_train.shape[0],\ "The size of X_train must be equals to the size of Y-Train" assert self.k <= X_train.shape[0] self._X_train = X_train self._Y_train = Y_train return self def predict(self,x_predict): return np.array([self._predict(x) for x in x_predict]) def _predict(self,x): distances = [ sqrt(np.sum((x_train-x)**2)) for x_train in self._X_train] nearest = np.argsort(distances) votes = [i for i in self._Y_train[nearest[:self.k]]] return Counter(votes).most_common(1)[0][0] def __repr__(self): return "knn(k=%d)" %self.k
测试与训练数据集分类
为了能够确认模型的准确性,我们需要将已有数据集按一定比例分类为测试数据集和训练数据集
# _*_ encoding:utf-8 _*_import numpy as npdef train_test_split(X,y,test_radio=0.2,seed=None): assert X.shape[0]==y.shape[0],"The size of X and y must be equal" assert 0.0<=test_radio<=1.0,"test radio must be valid" if(seed): np.random.seed(seed) shuffled_indexes = np.random.permutation(len(X)) test_size = int(X.shape[0]*test_radio) test_indexes = shuffled_indexes[:test_size] train_indexes = shuffled_indexes[test_size:] X_test = X[test_indexes] y_test = y[test_indexes] X_train = X[train_indexes] y_train = y[train_indexes] return X_train,X_test,y_train,y_test
使用knn算法测试数据集digits
import numpy as npfrom sklearn import datasetsimport matplotlib.pyplot as pltimport matplotlib %run MyScripts/KNN.py %run MyScripts/metrics.py %run MyScripts/model_selection.py digits = datasets.load_digits() X = digits.data y = digits.target
some_digit = X[666] some_digit_image = some_digit.reshape(8,8) plt.imshow(some_digit_image,cmap=matplotlib.cm.binary)
画出第666个数据对应的数字图片
knn_clf = KNNClassifier(k=6) X_train,X_test,y_train,y_test = train_test_split(X,y) knn_clf.fit(X_train,y_train) knn_clf.score(X_test,y_test)
超参数
超参数是模型运行前必须要决定的参数,例如k近邻算法中的k值和距离
确定超参数一般使用的方法:领域知识
经验数值
实验探索
确定knn算法用于digits数据集的最佳超参数
//使用网格搜索法确定weights和k超参数 best_k = -1 best_score = -1 methods = ["uniform","distance"] best_method = ""for method in methods: for k in range(1,11): knn_clf = KNeighborsClassifier(n_neighbors=k,weights=method) knn_clf.fit(X_train,y_train) score = knn_clf.score(X_test,y_test) if(score>best_score): best_k = k best_score = score best_method = methodprint("best_k = ",best_k)print("best_score = ",best_score)print("best_method = ",best_method)
best_k = 3
best_score = 0.9888888888888889
best_method = uniform
best_k = -1 best_score = -1 best_p=-1for p in range(1,6): for k in range(1,11): knn_clf = KNeighborsClassifier(n_neighbors=k,weights="distance",p=p) knn_clf.fit(X_train,y_train) score = knn_clf.score(X_test,y_test) if(score>best_score): best_k = k best_score = score best_p = pprint("best_k = ",best_k)print("best_score = ",best_score)print("best_p = ",best_p)
best_k = 3
best_score = 0.9888888888888889
best_p = 2
作者:冰源_63ad
链接:https://www.jianshu.com/p/9be4df54e32e