Open Access

Feature selection for high-dimensional data based on scaled cross operator threshold filtering specific memory algorithm

, , ,  and   
Mar 26, 2025

Cite
Download Cover

Figure 1.

General process of feature selection
General process of feature selection

Figure 2.

Basic flow of genetic algorithm
Basic flow of genetic algorithm

Figure 3.

Roulette selects cross individuals
Roulette selects cross individuals

Figure 4.

Running time (seconds) of the algorithm on five base BNS
Running time (seconds) of the algorithm on five base BNS

Figure 5.

Box diagram of feature dimensions obtained by four operators
Box diagram of feature dimensions obtained by four operators

Evaluation measures on different benchmarks BNs

Network Algorithm F1 Precision Recall Run-Time (s)
Child IAMB 0.79/0.68 0.87/0.68 0.78/0.79 0.09
MMMB 0.75/0.69 0.88/0.78 0.72/0.71 0.25
STMB 0.74/0.71 0.92/0.86 0.66/0.66 0.15
HITON-MB 0.83/0.82 0.95/0.85 0.76/0.76 0.25
WTMB 0.88/0.86 0.98/0.95 0.78/0.78 0.13
Child10 IAMB 0.63/0.63 0.61/0.62 0.66/0.65 1.18
MMMB 0.68/0.67 0.71/0.68 0.64/0.65 2.41
STMB 0.57/0.51 0.71/0.57 0.61/0.56 2.26
HITON-MB 0.78/0.73 0.91/0.76 0.72/0.74 2.31
WTMB 0.73/0.71 0.84/0.76 0.65/0.61 2.22
Pig IAMB 0.73/0.74 0.77/0.81 0.71/0.73 11.06
MMMB 0.82/0.81 0.81/0.81 0.82/0.82 13.01
STMB 0.76/0.72 0.76/0.82 0.74/0.70 11.06
HITON-MB 0.92/0.74 0.86/0.62 0.91/0.91 14.77
WTMB 0.93/0.92 0.92/0.92 0.93/0.90 11.76
Gene IAMB 0.62/0.67 0.61/0.63 0.67/0.67 20.08
MMMB 0.71/0.71 0.67/0.67 0.77/0.77 23.16
STMB 0.55/0.17 0.43/0.11 0.72/0.72 30.23
HITON-MB 0.81/0.79 0.78/0.67 0.91/0.90 23.33
WTMB 0.91/0.91 0.91/0.88 0.94/0.94 21.86
Hailfinder IAMB 0.21/0.22 0.24/0.23 0.16/0.17 0.21
MMMB 0.22/0.23 0.28/0.27 0.16/0.17 0.31
STMB 0.17/0.16 0.27/0.24 0.13/0.12 1.52
HITON-MB 0.21/0.21 0.31/0.27 0.16/0.16 0.76
WTMB 0.22/0.22 0.32/0.30 0.17/0.18 0.22

Experimental parameter Settings

Algorithm Parameters
IAMB Attenuation factor γ=0.2
MMMB Mutation probability=0.2,Mutation retry number=2
STMB c1=c2=3,ω=1.0124,θ=0.5,α=k/WM_rate
HITON-MB pmin=0.16,CR=0.3,F=0.4,α=0.4,β=0.3,σ=0.2

Benchmark BN’s (synthetic dataset)

Bayesian Network #Features #Edges Max In/Out Degree Min/Max |PC set|
Child 30 45 3/7 2/8
Child10 204 128 2/7 2/9
Pig 445 591 4/39 2/41
Gene 802 973 3/10 1/11
Hailfinder 58 65 5/16 2/17

Comparison of nonparametric effect sizes using Cliff’s delta

Dataset Algorithm δ
Child IAMB 0.76
MMMB 0.39
STMB 0.61
HITON-MB 0.75
WTMB 0.80
Child10 IAMB 0.75
MMMB 0.43
STMB 0.58
HITON-MB 0.71
WTMB 0.75
Pig IAMB 0.17
MMMB 0.39
STMB 0.42
HITON-MB 0.51
WTMB 0.49
Gene IAMB 0.33
MMMB 0.54
STMB 0.41
HITON-MB 0.19
WTMB 0.40
Hailfinder IAMB 0.61
MMMB 0.42
STMB 0.37
HITON-MB 0.55
WTMB 0.76
Language:
English