Evaluation of DBSCAN algorithm on different programming languages: An exploratory study

DBSCAN is a well-known clustering algorithm that is often used to find associations and structures in large spatial data. Due to its popularity, built-in functions for DBSCAN have been implemented on top of many different programming languages. Researchers and practitioners (i.e., data scientists) have been using these built-in functions to cluster and analyze a prolific area of research in data science. Due to the many implementations of DBSCAN and its utilization in many different languages, the output of each built-in DBSCAN function is assumed to be identical. In this paper, we present a systematic approach to evaluate the built-in functions of DBSCAN algorithms and to identify discrepancies in their output. The evidence from the study shows that there are some discrepancies and recommends caution in dealing with built-in functions.