generated from LufsX/Challenge-kNN
This commit is contained in:
+20
-112
@@ -1,96 +1,48 @@
|
||||
import numpy as np
|
||||
from sklearn.decomposition import PCA
|
||||
from sklearn.model_selection import cross_val_score, KFold
|
||||
from sklearn.metrics import accuracy_score
|
||||
from sklearn.preprocessing import StandardScaler
|
||||
import os
|
||||
|
||||
"""
|
||||
请在这里填写学号和姓名，不填写即成绩无效。
|
||||
"""
|
||||
id = "202210423103"
|
||||
name = "王坤"
|
||||
|
||||
def classify0(input_vector, dataSet, labels, k=3, weighted=False, distance_metric='euclidean'):
|
||||
def classify0(input_vector, dataSet, labels, k=3):
|
||||
"""
|
||||
KNN 分类算法（支持加权KNN、不同距离度量以及降维处理）。
|
||||
KNN 分类算法。
|
||||
|
||||
Args:
|
||||
- input_vector (numpy array): 输入待分类的样本。
|
||||
- dataSet (numpy array): 训练数据集。
|
||||
- labels (list or numpy array): 训练集对应的标签。
|
||||
- k (int): kNN 算法参数。
|
||||
- weighted (bool): 是否使用加权KNN。
|
||||
- distance_metric (str): 距离度量方法，支持 'euclidean', 'manhattan', 'minkowski'。
|
||||
- k (int): kNN 算法参数，请**自行设置默认值**。
|
||||
|
||||
Returns:
|
||||
- int: 预测的分类标签。
|
||||
"""
|
||||
# 根据选择的距离度量计算距离
|
||||
if distance_metric == 'euclidean':
|
||||
distances = np.linalg.norm(dataSet - input_vector, axis=1)
|
||||
elif distance_metric == 'manhattan':
|
||||
distances = np.sum(np.abs(dataSet - input_vector), axis=1)
|
||||
elif distance_metric == 'minkowski':
|
||||
p = 3 # 可以调整为合适的 p 值
|
||||
distances = np.sum(np.abs(dataSet - input_vector) ** p, axis=1) ** (1 / p)
|
||||
else:
|
||||
raise ValueError("Unsupported distance metric")
|
||||
# 计算输入向量与训练集中每个样本的欧氏距离
|
||||
distances = np.linalg.norm(dataSet - input_vector, axis=1)
|
||||
|
||||
# 按距离从小到大排序并获取前k个距离最小的索引
|
||||
# 按照距离从小到大排序并获取前k个距离最小的索引
|
||||
sorted_indices = distances.argsort()[:k]
|
||||
|
||||
# 如果加权，使用距离的倒数作为权重
|
||||
if weighted:
|
||||
weights = 1 / (distances[sorted_indices] + 1e-5) # 防止除零
|
||||
weighted_labels = {}
|
||||
for i, index in enumerate(sorted_indices):
|
||||
label = labels[index]
|
||||
weighted_labels[label] = weighted_labels.get(label, 0) + weights[i]
|
||||
predicted_label = max(weighted_labels, key=weighted_labels.get)
|
||||
else:
|
||||
# 统计前k个最近邻的标签
|
||||
nearest_labels = [labels[i] for i in sorted_indices]
|
||||
predicted_label = max(set(nearest_labels), key=nearest_labels.count)
|
||||
# 统计前k个最近邻的标签
|
||||
nearest_labels = [labels[i] for i in sorted_indices]
|
||||
|
||||
# 返回出现次数最多的标签
|
||||
predicted_label = max(set(nearest_labels), key=nearest_labels.count)
|
||||
|
||||
return predicted_label
|
||||
|
||||
def pca_reduction(data, n_components=2):
|
||||
"""
|
||||
使用PCA对数据进行降维。
|
||||
|
||||
Args:
|
||||
- data (numpy array): 原始特征数据。
|
||||
- n_components (int): 保留的主成分数量。
|
||||
|
||||
Returns:
|
||||
- numpy array: 降维后的特征数据。
|
||||
"""
|
||||
pca = PCA(n_components=n_components)
|
||||
reduced_data = pca.fit_transform(data)
|
||||
return reduced_data
|
||||
|
||||
def normalize_data(data):
|
||||
"""
|
||||
归一化特征数据，将每个特征值缩放到 [0, 1] 之间。
|
||||
|
||||
Args:
|
||||
- data (numpy array): 原始特征数据。
|
||||
|
||||
Returns:
|
||||
- numpy array: 归一化后的特征数据。
|
||||
"""
|
||||
min_vals = data.min(0)
|
||||
max_vals = data.max(0)
|
||||
ranges = max_vals - min_vals
|
||||
norm_data = (data - min_vals) / ranges
|
||||
return norm_data, ranges, min_vals
|
||||
|
||||
def load_data(file_path):
|
||||
"""
|
||||
从 TSV 文件中加载数据集，并进行归一化处理。
|
||||
从 TSV 文件中加载数据集。
|
||||
|
||||
Args:
|
||||
- file_path (str): TSV 文件路径。
|
||||
|
||||
Returns:
|
||||
- np.array: 归一化后的特征数据。
|
||||
- np.array: 特征数据。
|
||||
- np.array: 标签数据。
|
||||
"""
|
||||
data = np.loadtxt(file_path, delimiter='\t')
|
||||
@@ -98,56 +50,12 @@ def load_data(file_path):
|
||||
labels = data[:, -1].astype(int)
|
||||
return data_set, labels
|
||||
|
||||
def cross_validate_knn(data_set, labels, k_range, weighted=False, distance_metric='euclidean'):
|
||||
"""
|
||||
使用交叉验证来选择最佳的K值。
|
||||
|
||||
Args:
|
||||
- data_set (numpy array): 训练数据集。
|
||||
- labels (numpy array): 训练集标签。
|
||||
- k_range (list): K值的候选范围。
|
||||
- weighted (bool): 是否使用加权KNN。
|
||||
- distance_metric (str): 距离度量方法。
|
||||
|
||||
Returns:
|
||||
- tuple: 最佳K值和对应的最高准确率。
|
||||
"""
|
||||
best_k = k_range[0]
|
||||
best_accuracy = 0
|
||||
for k in k_range:
|
||||
kf = KFold(n_splits=5)
|
||||
accuracies = []
|
||||
for train_index, test_index in kf.split(data_set):
|
||||
train_data, test_data = data_set[train_index], data_set[test_index]
|
||||
train_labels, test_labels = labels[train_index], labels[test_index]
|
||||
predictions = [classify0(test_sample, train_data, train_labels, k, weighted, distance_metric)
|
||||
for test_sample in test_data]
|
||||
accuracy = accuracy_score(test_labels, predictions)
|
||||
accuracies.append(accuracy)
|
||||
mean_accuracy = np.mean(accuracies)
|
||||
if mean_accuracy > best_accuracy:
|
||||
best_accuracy = mean_accuracy
|
||||
best_k = k
|
||||
return best_k, best_accuracy
|
||||
|
||||
if __name__ == "__main__":
|
||||
# 加载数据集
|
||||
# 以下代码可用于测试，可自行随意修改
|
||||
train_file = "datingTrainSet2.tsv"
|
||||
train_data, train_labels = load_data(train_file)
|
||||
|
||||
# 归一化和PCA降维处理
|
||||
scaler = StandardScaler()
|
||||
scaled_train_data = scaler.fit_transform(train_data)
|
||||
pca_train_data = pca_reduction(scaled_train_data, n_components=3)
|
||||
|
||||
# 使用交叉验证选择最佳K值
|
||||
k_range = range(1, 21)
|
||||
best_k, best_accuracy = cross_validate_knn(pca_train_data, train_labels, k_range, weighted=True, distance_metric='euclidean')
|
||||
print(f"最佳K值: {best_k}, 交叉验证准确率: {best_accuracy}")
|
||||
|
||||
# 使用最佳K值进行预测
|
||||
# 示例测试样本，确保维度与训练数据一致
|
||||
test_sample = np.array([65915, 3.868359, 0.128078])
|
||||
norm_test_sample = scaler.transform([test_sample])
|
||||
pca_test_sample = pca_reduction(norm_test_sample, n_components=3)
|
||||
prediction = classify0(pca_test_sample[0], pca_train_data, train_labels, k=best_k, weighted=True)
|
||||
prediction = classify0(test_sample, train_data, train_labels)
|
||||
print(f"预测的分类标签: {prediction}")
|
||||
Reference in New Issue
Block a user