Game Data Mining: Clustering and Visualization of Online Game Data in Cyber-Physical Worlds

Since its debut in May 2016, Overwatch has quickly become a popular team-based online video game. Despite the popularity of Overwatch, many new players - who join the game unsure how to compete with the game's veterans - feel overwhelmed with the vast knowledge required to properly play at higher skill levels. In this paper, a data mining algorithm is designed and developed for clustering and visualization of online game data at the cyber-physical world boundary. Scientifically, the algorithm uses affinity propagation for clustering and two-dimensional graphs for visualizing online game data. The algorithm analyzes the Overwatch game data for the discovery of new knowledge about current players and the clustering of data for each hero character. This knowledge enables the analysis of individual clusters and provides statistics that have a high correlation with winning player strategies. These statistics are expected to have a large influence on how a character is played, and thus can aid new players in learning their priorities as each hero character. In other words, the algorithm helps analyze the online game playing data, get insight about the grouping or clusters of players, and offer suggestions to new players of the game.