A Novel Fuzzy K-Means Clustering Approach Optimized by Bacterial Foraging Algorithm for Document Categorization

Document categorization is a crucial task in organizing large collections of text. Traditional clustering methods like K-means often struggle with uncertainties in data. This paper presents a novel approach that combines Fuzzy K-Means (FKM) clustering with Bacterial Foraging Optimization (BFO) to enhance document clustering performance. The proposed method, FKM-BFO, benefits from fuzzy clustering’s ability to assign documents to multiple clusters, reflecting the inherent overlap in topics, while using the BFO algorithm to optimize the clustering process. FKM allows documents to belong to multiple clusters with varying degrees of membership, making it more suitable for real-world text data. However, FKM is sensitive to initial centroid placements and may get stuck in local optima. To address this, BFO, inspired by the foraging behaviour of bacteria, is used to optimize the initial centroids and guide the FKM algorithm to a global optimum. This combination improves clustering accuracy by better determining the cluster center and membership values. We evaluate the FKM-BFO approach using benchmark datasets like 20 Newsgroups and Reuters-21578. The results show that FKM-BFO outperforms traditional clustering methods, such as K-means and Fuzzy C-Means, in terms of accuracy and robustness, especially in handling noisy and high-dimensional data. This hybrid approach offers an effective solution for document categorization, providing higher accuracy and stability. Future work could explore its scalability and applicability to larger, real-time document clustering tasks.