Efficient clustering in data mining applications based on harmony search and k-medoids

There are now ever-increasing amounts of digital data stored daily. Processing such big data requires new sciences such as data mining. Data mining aims to find the information and knowledge hidden in these massive amounts of data. Data mining has various branches and applications, one of which is clustering. In clustering, data are classified into several clusters. The data of one cluster are similar to each other but different from those of other clusters. Processing time is considered a major challenge, given the use of big data in data mining applications. Metaheuristic algorithms can solve large-scale time-consuming problems in acceptable durations. The harmony search is one of such algorithms. This paper proposes a hybrid method for data clustering by integrating the harmony search algorithm (HSA) with k-medoids, which is a clustering algorithm. Based on k-medoids, the proposed method uses real data samples as cluster centroids to eliminate the effects of noise data. Furthermore, the HSA reduces the data clustering duration. In addition, trees social relations optimization algorithm (TSR) was used to increase the quality of the solutions in the proposed method. The proposed method was compared practically with the other techniques. According to the comparison results, the proposed method outperformed similar techniques in terms of the solution fitness and Dunn and Silhouette criteria. The temporal outputs of the proposed method were also acceptable. The best results obtained for the proposed method in the form of fitness of the final solution is equal to 1.02, execution time is equal to 2.9562 s, Dunn index is equal to 0.067 and Silhouette is equal to 1. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024.