An efficient approach to Data clustering using the K-Means algorithm in Big data analytics

Abstract

With the exponential growth of data, efficient clustering techniques are essential for extracting meaningful patterns in Big Data Analytics. The K-Means algorithm is widely used due to its simplicity and scalability. However, its performance is often hindered by high-dimensional data, initialization sensitivity, and computational complexity. This study proposes an optimized K-Means clustering approach that integrates an improved centroid initialization method and parallel processing to enhance efficiency in Big Data environments. The proposed method was evaluated using real-world datasets such as KDD Cup and UCI Machine Learning Repository, with data sizes ranging from 10GB to 100GB. Experimental results demonstrate a 30% reduction in execution time and a 15% improvement in clustering accuracy compared to traditional K-Means. The optimized approach also shows a 20% lower convergence time, making it suitable for large-scale applications. In conclusion, the enhanced K-Means algorithm significantly improves clustering performance in Big Data settings. The combination of advanced initialization and parallel computing ensures better scalability and accuracy, making it a viable solution for real-time analytics. Future work will focus on extending this approach to handle streaming data and non-Euclidean spaces.