Cleaning, analysing and visualising keyword co-occurrence, citation networks, species interaction networks and other complex relational data.
Prepare node and edge tables, unify names, remove duplicates, and handle synonyms and missing values.
The most common problem in network analysis is the same entity appearing under several names.
Import nodes and edges, and check whether the network is connected and whether there are isolated nodes or anomalous edges.
Check the data before adjusting the layout, or the graph can easily mislead.
Compute degree centrality, betweenness centrality, modularity, clustering coefficient and other metrics.
Choose metrics that serve the research question rather than piling up network jargon.
Set colours, sizes, labels and layout algorithms according to node category, weight and module.
Don't show every label; showing only key nodes is clearer.
After exporting SVG/PDF, optimise fonts, colours, legend and layout.
Network graphs easily become over-complex; in a paper, highlight the main structure and key nodes.
