<p>In the context of data clustering, this study focused on the crucial yet sometimes neglected issue of centroid identification. More specifically, the research was conducted within natural language processing applications to achieve two high-quality clusters. By utilizing traditional statistical measures of central tendency in conjunction with distance metrics that take into consideration centroid location, we developed and assessed a framework for selecting centroids. We established the viability of alternative centroid calculation methods without incurring complex computational overhead, which is a departure from the traditional approach that is based on the mean. Next, we evaluated the quality of the clusters by employing the Dunn index and the within-cluster sum of squares (WCSS). This was accomplished by conducting empirical research to investigate a variety of distance computation methods and the impact that these methods had on the selection of the centroid. The findings that we have obtained highlight the enormous influence that centroid selection has on the effectiveness of clustering. These findings provide useful insights that may be applied to better data analysis and machine learning applications in high-dimensional spaces. The purpose of this case study is to provide tangible recommendations for the development of future clustering algorithms that will improve the quality of clusters and the computation of centroid values.</p>