add KNN

Author: codewithzichao

KNN/KNN_python.py

@@ -0,0 +1,113 @@
+#coding:utf-8
+#Author:codewithzichao
+#E-mail:[email protected]
+
+'''
+数据集：Mnist数据集(只使用了1000来训练，只使用了1000来测试。)
+结果(准确率)：0.738
+时间：28.6643168926239
+'''
+import  numpy as np
+import time
+
+def loadData(fileName):
+    '''
+    加载数据
+    :param fileName: 数据路径
+    :return: 返回特征向量与标签类别
+    '''
+    data_list=[]
+    label_list=[]
+
+    with open(fileName,"r") as f:
+        for line in f.readlines():
+            curline=line.strip().split(",")
+
+            data_list.append([int(feature) for feature in curline[1:]])
+            label_list.append(int(curline[0]))
+
+        data_matrix=np.array(data_list)
+        label_matrix=np.array(label_list)
+
+        return data_matrix,label_matrix
+
+class KNN(object):
+    def __init__(self,train_data,train_label,K):
+        '''
+        构造函数
+        :param train_data: 训练集的特征向量
+        :param train_label: 训练集的类别向量
+        :param K: 指定的K值
+        '''
+        self.train_data=train_data
+        self.train_label=train_label
+        self.input_num=self.train_data.shape[0]
+        self.feature=self.train_data.shape[1]
+        self.K=K
+
+    def cal_distance(self,x1,x2):
+        '''
+        计算两个样本之间的距离，使用欧式距离
+        :param x1: 第一个样本
+        :param x2: 第二步样本
+        :return: 样本之间的距离
+        '''
+        return np.sqrt(np.sum(np.square(x1-x2)))
+
+    def get_K(self,x):
+        dist_group=np.zeros(self.input_num)
+        for i in range(self.input_num):
+            x1=self.train_data[i]
+            dist=self.cal_distance(x,x1)
+            dist_group[i]=dist
+
+        topK=np.argsort(dist_group)[:self.K]#升序排序
+
+        labeldist=np.zeros(10)#10个标签，在每一个标签对应的位置上加1
+
+        for i in range(len(topK)):
+            labeldist[int(self.train_label[topK[i]])]+=1
+
+        return np.argmax(labeldist)
+
+    def test(self,test_data,test_label):
+        '''
+        在测试集上测试
+        :param test_data: 测试集的特征向量
+        :param test_label: 测试集的标签向量
+        :return: 准确率
+        '''
+        error=0
+
+        test_num=test_data.shape[0]
+        for i in range(test_num):
+            print(f"the current sample is {i+1},the total samples is{test_num}.")
+            x=test_data[i]
+            y=self.get_K(x)
+
+            if(y!=test_label[i]):
+                error+=1
+
+        accuracy=(test_num-error)/test_num
+        return accuracy
+
+if __name__=="__main__":
+    start=time.time()
+
+    print("start load data.")
+    train_data,train_label=loadData("../MnistData/mnist_train.csv")
+    test_data,test_label=loadData("../MnistData/mnist_test.csv")
+    print("finished load data.")
+
+    a=KNN(train_data[:1000],train_label[:1000],30)
+
+    print("finished training.")
+
+    accuracy=a.test(test_data[:1000],test_label[:1000])
+    print(f"the accuracy is {accuracy}.")
+
+    end=time.time()
+
+    print(f"the total time is {end-start}.")
+
+

KNN/KNN_sklearn.py

@@ -0,0 +1,59 @@
+#coding:utf-8
+#Author:codewithzichao
+#E-mail:[email protected]
+
+'''
+数据集：Mnist数据集(只使用了1000来训练，只使用了1000来测试。)
+结果(准确率)：0.799
+时间：16.832828998565674
+---------------------------
+果然，自己写的python没有编写kdtree等部分，效果与时间上都比不上sklearn。
+'''
+
+import numpy as np
+import time
+from sklearn.neighbors import KNeighborsClassifier
+
+def loadData(fileName):
+    '''
+    加载数据
+    :param fileName: 数据路径
+    :return: 返回特征向量与标签类别
+    '''
+    data_list=[]
+    label_list=[]
+
+    with open(fileName,"r") as f:
+        for line in f.readlines():
+            curline=line.strip().split(",")
+
+            data_list.append([int(feature) for feature in curline[1:]])
+            label_list.append(int(curline[0]))
+
+        data_matrix=np.array(data_list)
+        label_matrix=np.array(label_list)
+
+        return data_matrix,label_matrix
+
+
+if __name__=="__main__":
+    start = time.time()
+
+    print("start load data.")
+    train_data, train_label = loadData("../MnistData/mnist_train.csv")
+    test_data, test_label = loadData("../MnistData/mnist_test.csv")
+    print("finished load data.")
+
+    knn=KNeighborsClassifier(n_neighbors=10)
+    knn.fit(train_data[:1000],train_label[:1000])
+
+    prediction=knn.predict(test_data[:1000])
+    for i in range(1000):
+        print(f"predict is {prediction[i]},the true is {test_label[i]}.")
+    accuracy=knn.score(test_data[:1000],test_label[:1000])
+    print(f"the accuracy is {accuracy}.")
+
+    end=time.time()
+
+    print(f"the total time is {end-start}.")
+