将数据集转为hdf5格式

安装 h5py

pip install h5py

转化hdf5格式

这里以 scene 数据集为例，该数据集可在 https://mulan.sourceforge.net/datasets-mlc.html 中获取

import h5py
from scipy.io import arff
import numpy as np

source_path = r'./scene.arff'
target_path = r'./scene.hdf5'

data, meta = arff.loadarff(source_path)
feature = []
target = []
for i in range(data.shape[0]):
	t = list(data[i])
	feature.append(t[:-6])
	target.append([int(bytes.decode(v)) for v in t[-6:]])
feature = np.array(feature)
target = np.array(target)

# 保存到 target_path
with h5py.File(target_path, 'w') as hf:
	hf.create_dataset('feature', data=feature)
	hf.create_dataset('target', data=target)
# 查看
with h5py.File(target_path, 'r') as hf:
	print('keys: ', hf.keys())
	print(hf['feature'].shape)
	print(hf['target'])

查看 HDF 文件

推荐使用 https://portal.hdfgroup.org/downloads/hdfview/hdfview3_3_2.html