Centroid-Guided Firefly Algorithm Improves Automatic Clustering
A new version of the Firefly Algorithm (FA) has been created to address the limitations of traditional clustering methods like K-Means, which struggle with uneven cluster shapes and densities and require users to specify the number of clusters in advance. This updated algorithm introduces a method for adjusting centroids and uses a multi-objective fitness function that considers compactness, separation, and a penalty related to TSP navigation. It can automatically find the best number of clusters and adjust their boundaries as necessary. Tests on robotic sensor networks showed that this approach improves clustering quality and reduces intra-cluster path lengths compared to K-Means, highlighting its potential for complex spatial clustering and future applications in higher-dimensional and adaptive scenarios.
Key facts
- The algorithm is a variant of the Firefly Algorithm (FA).
- It addresses limitations of K-Means with non-uniform cluster shapes and densities.
- It eliminates the need to pre-define the number of clusters.
- It uses a centroid movement strategy and multi-objective fitness function.
- The fitness function includes compactness, separation, and TSP-based navigation penalty.
- It automatically estimates the optimal number of clusters.
- Experiments show improved clustering quality and reduced intra-cluster path distances.
- Application to robotic sensor networks highlights practical value.
Entities
Institutions
- arXiv