Development and Evaluation of a Parallel K-means Algorithm for Big Data Analysis in Google MapReduce Environment

The challenge of executing iterative big data analysis algorithms within the Google Cloud MapReduce environment has been addressed by developing a parallel K-means algorithm capable of leveraging the distributed computing power of the platform.Traditional K-means, which requires iterative steps, is adapted into a parallel version using MapReduce to enhance computational efficiency.This parallel algorithm is structured into multiple super-steps, each of which executes in parallel but is processed sequentially across super-steps.Each super-step corresponds to one iteration of the serial K-means algorithm, with parallel computation carried out at each node to determine the mean of each cluster center.Experimental evaluations have demonstrated that the parallel K-means algorithm performs effectively and accurately.Notably, for a dataset of 450 water samples, a parallel speedup factor of 20.8 was achieved, significantly reducing the time required for data analysis.This substantial reduction in processing time is critical in time-sensitive applications, such as coal mine rescue operations, where quick decision-making is essential.The results indicate that the proposed parallel K-means algorithm is both a feasible and efficient solution for handling large-scale datasets within cloud environments, providing substantial benefits in both computational speed and practical application.