Straight line, 4D AND, 5D XOR, Hypersphere, Cone, Trigonometric, Double Spiral, Yin-yang, 5 multi-cut, 10 multi-cut
收藏资源简介:
用于基准测试特征选择算法的合成数据集,包括多种类型的数据集,如直线、4D AND、5D XOR等,每个数据集都有其特定的特征和目标变量方程。
A synthetic dataset for benchmarking feature selection algorithms, encompassing various types of datasets such as linear, 4D AND, 5D XOR, etc. Each dataset is characterized by its specific features and target variable equations.
数据集概述
数据集列表
| 数据集名称 | 相关特征数 | 冗余特征数 | 无关特征数 | 实例数 | 目标变量方程 |
|---|---|---|---|---|---|
| Straight line | 2 | 20 | 100 | 200 | F2 < F1 |
| 4D AND | 4 | 8 | 100 | 100 | x1 · x2 + x3 · x4 |
| 5D XOR | 5 | 10 | 100 | 100 | x1 ⊕ x2 ⊕ x3 ⊕ x4 ⊕ x5 |
| Hypersphere | 3 | 20 | 100 | 400 | 30^2 > F1^2 + F2^2 + F3^2 |
| Cone | 3 | 20 | 100 | 400 | F12 + F22 > F32 |
| Trigonometric | 2 | 5 | 50 | 200 | F2 < 5 sin(F1) |
| Double Spiral | 3 | 30 | 120 | 200 | 详见README |
| Yin-yang | 2 | 10 | 75 | 600 | - |
| 5 multi-cut | 6 | 20 | 100 | 500 | 详见README |
| 10 multi-cut | 6 | 20 | 100 | 500 | 详见README |
数据集加载方法
python from Synth_Select import load_dataset df = load_dataset(name)
可用数据集名称包括:
- 10_class_multicut
- 4D_AND
- 5_class_multicut
- 5D_XOR
- cone
- double_spiral
- hypersphere_3D
- trig
- y=x
- yinyang
数据集扩展方法
添加数值型无关特征
python from Synth_Select import addIrrelevant df = addIrrelevant(df, numIrrelevantCols=5, distributions=[normal, gamma, exponential])
添加数值型冗余特征
python from Synth_Select import addRedundant df = addRedundant(df, numRedFeats=5)
添加类别型无关特征
python from Synth_Select import addIrrelevantCat df = addIrrelevantCat(df, numIrrelevantCols=5, distributions=[normal, gamma, exponential])
添加类别型冗余特征
python from Synth_Select import addRedundantCat df = addRedundantCat(df, numRedFeats=5)
分布类型包括:
- "normal"
- "gamma"
- "exponential"




