Measure of Node Similarity in Multilayer Networks

Area of Science:

• Social Network Analysis
• Computational Social Science
• Network Science

Background:

• Link weights in networks often reflect node similarity.
• Analyzing node similarity across varying link weights requires robust measures.
• Understanding social homophily is crucial in social network analysis.

Purpose of the Study:

• Introduce a tunable measure for analyzing node similarity across different link weights.
• Apply this measure to study homophily in a freshman student population.
• Investigate the relationship between social connections, similarity, and network layers.

Main Methods:

• Developed a novel tunable measure for node similarity in weighted networks.
• Collected data from 659 university freshmen using smartphones and questionnaires.
• Constructed a weighted multilayer network from telecommunication and face-to-face contact data.

Main Results:

• No significant similarity in basic personality traits was found between strongly connected individuals.
• Socio-demographic variables exhibited significant similarity among connected individuals.
• Similarity varied across network layers; gender showed similarity at low weights and dis-similarity at high weights.

Conclusions:

• Node similarity is not solely determined by connection strength; link weight and network layer are critical.
• Socio-demographic factors play a more significant role in homophily than personality traits in this context.
• The developed measure effectively reveals nuanced patterns of similarity and dis-similarity within multilayer social networks.