机器学习+KNN+代码实现

import numpy import pandas from sklearn.datasets import load_iris import matplotlib.pyplot as plt # 加载Iris数据集，返回data和m、n # data结构说明：前四列为样本属性值，最后一列为样本标签 def LoadData(): Data = load_iris() m, n = Data.data.shape # 获取数据集样本数m、属性数n x = numpy.array(Data['data']) y = numpy.array(Data['target']) data = numpy.column_stack([x, y]) # 将x、y横向拼接 print('---------This is output of LoadData----------') # print(Data.DESCR) # 打印数据集说明 print(data) return data, m, n # 计算欧氏距离并返回 def Edist(x_training, row): dist = numpy.sqrt(numpy.sum(numpy.power(x_training - row, 1))) return dist # 利用numpy库实现样本随机抽样，划分出训练集和测试集 def Divide_Dataset(data, m): Index = numpy.random.choice(range(1, 150), m // 5, replace=False) # 不返回随机抽样 test = data[Index, :] train = [] for i in range(m): if i != Index.all(): # 剩余样本放入训练集 train.append(data[i]) train = numpy.array(train) return test, train # 训练KNN模型并预测测试集 def MyKNN(test, train, K): predicts = [] TrueCount = 0 for i in range(test.shape[0]): arr_dist = numpy.zeros(shape=(train.shape[0], 2)) for j in range(train.shape[0]): dist = Edist(test[i, :-1], train[j, :-1]) # 调用Edist函数计算样本点间欧氏距离 arr_dist[j, :] = dist, train[j, -1] df = pandas.DataFrame(data=arr_dist, columns=['dist', 'target']) mode = df.sort_values(by='dist')['target'].head(K).mode()[0] # 按照距离大小排序样本 predicts.append(mode) # 获得预测结果 if mode == test[i, -1]: # 统计测试集预测结果正确个数 TrueCount += 1 accuracy = TrueCount / test.shape[0] return accuracy, predicts if __name__ == '__main__': data, m, n = LoadData() test, train = Divide_Dataset(data, m) Accuracies = [] Predicts = [] for K in range(2, 10): # 迭代K值 accuracy, predict = MyKNN(test, train, K) Accuracies.append(accuracy) Predicts.append(predict) print('The predict result:\n', numpy.array(Predicts)) print('The Label of testset:\n', test[:, -1]) print('\nThe accuracy for k in 2-10:\n', numpy.array(Accuracies)) plt.plot(range(2, 10), Accuracies, label='K') plt.legend() plt.show()