Flexible optimization of hierarchical graph layout by genetic algorithm with various conditions

Graph layouts visualize relationships among data entities, where nodes represent individual entities and edges represent their relationships. Hierarchical graph layouts efficiently provide an overview of large-scale graphs, where nodes form clusters (called metanodes in this paper) based on their properties. Here, it is challenging to determine layouts for large-scale graphs, particularly hierarchical ones. Although various graph layout drawing methods, such as force-directed layout, have been discussed so far, the quality of a layout heavily relies on the initial positions of nodes or metanodes. Furthermore, it is more challenging to obtain layouts where specific desired metanodes stand out. This paper presents a layout optimization method for hierarchical graphs using a genetic algorithm (GA). Our method allows for the consistent improvement of layouts compared to relying solely on an existing algorithm for generating hierarchical graph layouts. In our implementation, first, hierarchical graph layouts are generated by applying an existing algorithm multiple times. Then, they are evaluated by specific metrics for hierarchical graph layouts and optimized using GA. Consequently, optimal layouts for these metrics are obtained. The paper also presents particular examples of layouts optimized under different conditions using a co-authorship graph dataset.