利用kmeans 对数据聚类

最新推荐文章于 2024-05-20 20:08:58 发布

苗妮

最新推荐文章于 2024-05-20 20:08:58 发布

阅读量118

点赞数

分类专栏：地震数据处理文章标签： kmeans 聚类算法

本文链接：https://blog.csdn.net/u014655960/article/details/131927365

版权

地震数据处理专栏收录该内容

10 篇文章 18 订阅

订阅专栏

import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans

from model.resnet import resnet18
import torch
from PIL import Image
from sklearn.decomposition import PCA


vmodel_path = "G:/minfan/myTask/MSTC_FWI/dataset/npy/SEG_trainData_130groups_1channels_201x301_vmodelData.npy"
vmodel = np.load(vmodel_path)

# Convert the NumPy array to a supported data type (e.g., float32)
vmodel = vmodel.astype(np.float32)

# Convert the NumPy array to a PyTorch tensor
vmodel_tensor = torch.tensor(vmodel)

vmodel_tensor = vmodel_tensor.reshape(vmodel_tensor.shape[0], 1, 201, 301)

# Using expand() method to convert to a 3-channel image  1通道扩展到3通道
expanded_vmodel_tensor = vmodel_tensor.expand(-1,  3, -1, -1)


# 加载预训练的ResNet模型，并移除最后的全连接分类层， 只提取特征
resnet = resnet18(pretrained=True)
resnet = torch.nn.Sequential(*list(resnet.children())[:-1])  # 移除最后一层  提取特征

resnet.eval()  # 设置为评估模式，这将冻结模型的参数

# 提取特征向量
with torch.no_grad():
    output = resnet(expanded_vmodel_tensor)

# 获取特征向量
feature_vector = output.squeeze().numpy()

# 输出特征向量的形状
print("Feature Vector Shape:", feature_vector.shape)


# 使用 K-Means 算法进行聚类
n_clusters = 2  # 要聚类的簇数
kmeans = KMeans(n_clusters=n_clusters, random_state=42)
kmeans.fit(feature_vector)
# 每个样本的聚类标签
labels = kmeans.labels_
print(labels)
# 聚类中心
cluster_centers = kmeans.cluster_centers_

'''

## 降维后显示方便可视化

# 使用PCA降维至2维
pca = PCA(n_components=2)
reduced_feature_vector = pca.fit_transform(feature_vector)

# 使用K-Means算法进行聚类
n_clusters = 2
kmeans = KMeans(n_clusters=n_clusters, random_state=42)
kmeans.fit(reduced_feature_vector)
labels = kmeans.labels_
cluster_centers = kmeans.cluster_centers_
'''


# 可视化聚类结果
plt.scatter(feature_vector[:, 0], feature_vector[:, 1], c=labels, s=50, cmap='viridis')
plt.scatter(cluster_centers[:, 0], cluster_centers[:, 1], c='red', marker='x', s=200, label='Centroids')
plt.xlabel('Feature 1')
plt.ylabel('Feature 2')
plt.title('K-Means Clustering')
plt.legend()
plt.show()