.png){kind=icon}
.png){kind=icon}
Return
Using a binary hybrid Fox optimization algorithm for attribute selection
.png){kind=icon}
PRE
AI
DOI：10.1007/s10586-025-05805-z.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The presence of irrelevant or redundant features in datasets impairs classification accuracy and increases data size, creating significant challenges for the development of intelligent information systems. To address these challenges, this research introduces the Binary Fox Optimization Algorithm (BFOX), a binary hybrid attribute selection method that integrates the Fox Optimizer (FOX) with chaotic maps. This method enhances feature selection efficiency by reducing dimensionality and improving classification accuracy, thereby supporting the design of next-generation intelligent information systems. The performance of BFOX was evaluated on ten benchmark datasets from the UCI repository. Its effectiveness was compared with several state-of-the-art methods reported in the literature, including recently developed algorithms such as BCDDO, GOOSE, LEO, and LPB. For the classification tasks, the system applied the Euclidean distance matrix and the K-Nearest Neighbors (K-NN) algorithm. Each experiment was repeated twenty times, and key metrics including the number of selected attributes, classification accuracy, and fitness were systematically assessed. The results demonstrated that BFOX achieved the highest mean classification accuracy on nine of the ten datasets. The only exception was BreastEW, where the LPB algorithm produced the best result, while BFOX delivered a very close and competitive performance. In addition, BFOX reduced the number of attributes in seven datasets (Zoo, WineEW, Lymphography, HeartEW, Breastcancer, BreastEW, SpectEW), confirming its effectiveness in generating compact feature subsets without compromising classification performance. Overall, the Binary Fox Optimization Algorithm provides an effective and reliable solution for attribute selection. It is directly applicable to data mining, knowledge discovery, and the development of intelligent information systems designed to process uncertainty and large-scale data. In summary, the findings confirm that the proposed BFOX algorithm consistently achieves superior classification accuracy and reduced feature dimensionality compared to existing methods. By combining the FOX optimizer with chaotic maps in a binary framework, BFOX offers a statistically validated, efficient, and reliable solution for feature selection tasks, supporting its practical applicability in intelligent information systems. Furthermore, the superiority of BFOX was statistically validated using the Friedman test, which confirmed that the observed performance differences among algorithms are significant.
Keywords:
Attribute selection
Fox
Classification
Chaotic maps
Optimization
Journal
C
IF:
0
Papers:
691
Citations:
1