generated from LufsX/Experiment-kNN
96 lines
2.6 KiB
Python
96 lines
2.6 KiB
Python
import numpy as np
|
|
import os
|
|
|
|
# 学号和姓名
|
|
id = "202210423118"
|
|
name = "刘国庆"
|
|
|
|
def classify(input_vector, dataSet, labels, k=3):
|
|
"""
|
|
kNN分类算法.
|
|
|
|
Args:
|
|
- input_vector: 输入待分类的样本.
|
|
- dataSet: 训练数据集.
|
|
- labels: 训练集对应的标签.
|
|
- k: kNN参数.
|
|
|
|
Returns:
|
|
- 预测的分类标签.
|
|
"""
|
|
# 计算欧氏距离
|
|
distances = np.linalg.norm(dataSet - input_vector, axis=1)
|
|
# 获取距离最近的k个样本
|
|
nearest_neighbors = np.argsort(distances)[:k]
|
|
# 统计每个标签的投票数
|
|
label_count = {}
|
|
for i in nearest_neighbors:
|
|
vote_label = labels[i]
|
|
label_count[vote_label] = label_count.get(vote_label, 0) + 1
|
|
# 返回投票最多的标签
|
|
return max(label_count, key=label_count.get)
|
|
|
|
def load_data(directory):
|
|
"""
|
|
加载手写数字数据集.
|
|
|
|
Args:
|
|
- directory: 数据集所在目录.
|
|
|
|
Returns:
|
|
- 特征数据集和标签集.
|
|
"""
|
|
labels = []
|
|
file_list = os.listdir(directory)
|
|
num_files = len(file_list)
|
|
dataset = np.zeros((num_files, 1024)) # 每个图像为32x32, 转换为1x1024向量
|
|
|
|
for i, filename in enumerate(file_list):
|
|
# 标签是文件名的第一个字符
|
|
label = int(filename.split('_')[0])
|
|
labels.append(label)
|
|
# 转换图像为向量
|
|
dataset[i, :] = img_to_vector(os.path.join(directory, filename))
|
|
|
|
return dataset, np.array(labels)
|
|
|
|
def img_to_vector(filename):
|
|
"""
|
|
将32x32的二值图像转换为1x1024向量.
|
|
|
|
Args:
|
|
- filename: 图像文件路径.
|
|
|
|
Returns:
|
|
- 向量化的图像数据.
|
|
"""
|
|
vector = np.zeros((1, 1024))
|
|
with open(filename) as file:
|
|
for i in range(32):
|
|
line = file.readline().strip()
|
|
for j in range(32):
|
|
vector[0, 32 * i + j] = int(line[j])
|
|
return vector
|
|
|
|
|
|
if __name__ == "__main__":
|
|
# 加载训练数据集
|
|
train_data, train_labels = load_data("trainingDigits") # 使用正确的函数名 load_data
|
|
|
|
# 加载测试数据集
|
|
test_data, test_labels = load_data("testDigits") # 加载测试集文件夹中的所有数据
|
|
|
|
# 测试集预测结果
|
|
correct_count = 0
|
|
total_count = len(test_labels)
|
|
|
|
for i in range(total_count):
|
|
predicted_label = classify(test_data[i], train_data, train_labels)
|
|
print(f"测试样本 {i+1} - 真实标签: {test_labels[i]}, 预测标签: {predicted_label}")
|
|
if predicted_label == test_labels[i]:
|
|
correct_count += 1
|
|
|
|
accuracy = correct_count / total_count
|
|
print(f"分类精度: {accuracy * 100:.2f}%")
|
|
|