Skip to main content

TableĀ 2 Summary of link features. We denote the Twitter neighbourhood as \(\pmb{\Gamma^{T}}\) and the Foursquare neighbourhood as \(\pmb{\Gamma^{F}}\)

From: A multilayer approach to multiplexity and link prediction in online geo-social networks

Twitter features

mentions

\(|\textit{mentions}_{ij}|\)

hashtags

\(|\textit{hashtags}_{ij}|\)

overlap

\(\frac{|\Gamma_{i}^{T} \cap \Gamma_{j}^{T}|}{|\Gamma_{i}^{T} \cup \Gamma_{j}^{T}|}\)

aa_sim

\(\sum_{z \in \Gamma_{i}^{T} \cap \Gamma_{j}^{T}} \frac{1}{\log(|\Gamma_{z}^{T}|)}\)

Foursquare features

colocs

\(|\textit{colocations}_{ij}|\)

dist

\(\textit{haversine}(\textit{lat}_{i},\textit{lon}_{i}, \textit{lat}_{j}, \textit{lon}_{j})\)

overlap

\(\frac{|\Gamma_{i}^{F} \cap \Gamma_{j}^{F}|}{|\Gamma_{i}^{F} \cup \Gamma_{j}^{F}|}\)

aa_sim

\(\sum_{z \in \Gamma_{i}^{F} \cap \Gamma_{j}^{F}} \frac{1}{\log(|\Gamma_{z}^{F}|)}\)

Multilayer features

\(\textit{int}_{GNij} \)

\(\sum_{\alpha}^{M} |int^{\alpha}_{ij}|\)

\({\textit{sim}}_{GNij}\)

\(\frac{\textit{sim}_{ij}^{a}}{\textit{dist}_{ij}^{b}}\)

overlap

\(\frac{|\Gamma_{CNi} \cap \Gamma_{CNj}|}{|\Gamma_{CNi} \cup \Gamma_{CNj}|}\)

aa_sim

\(\sum_{z \in \Gamma_{CNi} \cap \Gamma_{CNj}} \frac{1}{\log(|\Gamma_{CNz}|)}\)