量化交易之机器学习篇 - 实现K近邻模型的两种方式

# 导入相关模块

import numpy as np

from collections import Counter
import matplotlib.pyplot as plt

from sklearn import datasets
from sklearn.utils import shuffle


def load_data():
    iris = datasets.load_iris()

    # 打乱数据后的数据和标签
    X, y = shuffle(iris.data, iris.target, random_state=13)

    # 数据转换为 flout32 格式
    X = X.astype(np.float32)

    # 简单划分训练集和测试集, 训练样本 - 测试样本比例为 7:3
    offset = int(X.shape[0] * 0.7)

    X_train, y_train = X[:offset], y[:offset]
    X_test, y_test = X[offset:], y[offset:]

    # 将标签转换为竖向量
    y_train = y_train.reshape((-1, 1))
    y_test = y_test.reshape((-1, 1))

    return X_train, X_test, y_train, y_test


def compute_distances(X, X_train):
    """
    定义欧氏距离函数
    X: 测试样本实例矩阵
    X_train: 训练样本实例矩阵
    """

    # 测试实例样本
    num_test = X.shape[0]

    # 训练实例样本量
    num_train = X_train.shape[0]

    # 基于训练和测试维度的欧氏距离初始化
    dists = np.zeros((num_test, num_train))

    # 测试样本鱼训练样本的矩阵点乘
    M = np.dot(X, X_train.T)

    # 测试样本矩阵平方
    te = np.square(X).sum(axis=1)

    # 训练样本矩阵平方
    tr = np.square(X_train).sum(axis=1)

    # 计算欧式距离
    dists = np.sqrt(-2 * M + tr + np.matrix(te).T)

    return dists

def predict_labels(y_train, dists, k=1):
    """
    定义预测函数
    :param y_train: 训练集标签
    :param dists: 测试集与训练集之间的欧氏距离矩阵
    :param k: k值
    :return: 测试集预测结果
    """

    # 测试样本量
    num_test = dists.shape[0]

    # 初始化测试集预测结果
    y_pred = np.zeros(num_test)

    # 遍历
    for i in range(num_test):
        # 初始化最近邻列表
        closest_y = []

        # 按 欧式距离矩阵排序后取索引, 并用训练集标签按排序后的索引取值

        # 最后展平列表
        # 注意 np.argsort 函数的用法
        labels = y_train[np.argsort(dists[i, :])].flatten()

        # 取最近的k个值
        closest_y = labels[0:k]
        # 对最近的k个值进行计数统计
        # 这里注意 collections 模块中的计数器 Counter 的用法
        c = Counter(closest_y)

        # 取计数最多的那个类别
        y_pred[i] = c.most_common(1)[0][0]

    return y_pred



if __name__ == '__main__':
    # 导入 sklearn iris 数据集
    X_train, X_test, y_train, y_test = load_data()

    dists = compute_distances(X=X_test, X_train=X_train)

    y_test_pred = predict_labels(y_train=y_train, dists=dists, k=1)
    y_test_pred = y_test_pred.reshape((-1, 1))

    # 找出预测正确的实例
    num_correct = np.sum(y_test_pred == y_test)

    # 计算分类准确率
    accuracy = float(num_correct) / X_test.shape[0]
    print('KNN Accuracy based on NumPy: ' + str(accuracy))


    # 用五折交叉验证寻找最优 k值
    # 五折
    num_folds = 5

    # 候选 k 值
    k_choices = [1, 3, 5, 8, 10, 12, 15, 20, 50, 100]
    X_train_folds = []
    y_train_folds = []

    # 训练数据划分
    X_train_folds = np.array_split(X_train, num_folds)

    # 训练标签划分
    y_train_folds = np.array_split(y_train, num_folds)
    k_to_accuracies = {}

    # 表里所有候选k值
    for k in k_choices:
        # 五折遍历
        for fold in range(num_folds):
            # 为 传入的训练集单独划分出一个验证集作为测试集
            validation_X_test = X_train_folds[fold]
            validation_y_test = y_train_folds[fold]
            temp_X_train = np.concatenate(X_train_folds[:fold] + X_train_folds[fold+1:])
            temp_y_train = np.concatenate(y_train_folds[:fold] + y_train_folds[fold+1:])

            # 计算距离
            temp_dists = compute_distances(X=validation_X_test, X_train=temp_X_train)
            temp_y_test_pred = predict_labels(temp_y_train, temp_dists, k=k)
            temp_y_test_pred = temp_y_test_pred.reshape((-1, 1))

            # 查看分类准确率
            num_correct = np.sum(temp_y_test_pred == validation_y_test)
            num_test = validation_X_test.shape[0]
            accuracy = float(num_correct) / num_test
            k_to_accuracies[k] = k_to_accuracies.get(k, []) + [accuracy]


    for k in sorted(k_to_accuracies):
        for accuracy in k_to_accuracies[k]:
            print(f'k = {k}, accuracy = {accuracy}')


    # 打印不同k值, 不同折数下的分类准确率
    for k in k_choices:
        # 取出第k个k值的分类准确率
        accuracies = k_to_accuracies[k]

        # 绘制不同k值下分类准确率的散点图
        plt.scatter([k] * len(accuracies), accuracies)

    # 计算分类准确率均值并排序
    accuracies_mean = np.array([np.mean(v) for k, v in sorted(k_to_accuracies.items())])
    # 计算分类准确率标准差并排序
    accuracies_std = np.array([np.std(v) for k, v in sorted(k_to_accuracies.items())])

    # 绘制有质询区间的误差棒图
    plt.errorbar(k_choices, accuracies_mean, yerr=accuracies_std)

    # 绘图标题
    plt.title('Cross-validation on k')

    # x轴标签
    plt.xlabel('k')

    # y轴标签
    plt.ylabel('Cross-validation accuracy')
    plt.show()

if __name__ == '__main__':
    # 导入 KNeighborsClassifier 模块
    from sklearn.neighbors import KNeighborsClassifier

    # 创建 k近邻实例
    neigh = KNeighborsClassifier(n_neighbors=10)

    # k 近邻模型拟合
    neigh.fit(X_train, y_train)

    # k 近邻模型预测
    y_pred = neigh.predict(X_test)

    # 预测结果数组重塑
    y_pred = y_pred.reshape((-1, 1))

    # 统计预测正确的个数
    num_correct = np.sum(y_pred == y_test)

    # 计算分类准确率
    accuracy = float(num_correct) / X_test.shape[0]
    print(f'KNN Accuracy based on sklearn: {accuracy}.')