A bimodal social network analysis to recommend points of interest to tourists

With the progressive role of computers and their users as actors in social networks, computations alike social network analysis (SNA) are gaining in attention. This work proposes an approach based on SNA, not alone on social networks as supported by existing approaches, to estimate the tourists’ satisfaction with individual Points of Interest (POIs), and accordingly recommend those POIs or not to that tourist or its tour planning system. Moreover, instead of a common unimodal network, a bimodal tourist–reviewer network is modeled as suggested by the SNA literature given tourists and POI reviewers act as two distinct classes of entities with links between them representing their (dis)similarities. Both tourists and reviewers provide their personal attributes (like age), but reviewers then providing preferences for specific POIs, whereas tourists only preferences for certain types or categories of POIs (say archeology). Further, an algorithm for grouping into “islands” of most similar reviewers to a certain tourist given the strength of corresponding links in the bimodal network is developed. Additionally, a ranking algorithm based on in-degree or authority centrality is adopted to identify the highest ranked reviewers within the island and recommend their preferred POIs to a given tourist. If there are more than single POIs preferred per reviewer, and there remain more than requested POIs of the highly ranked reviewers to select among for recommendation, a similar centrality algorithm is applied over a reviewer–POI network with links representing a certain reviewer prefers that certain POI. The evaluation initially with an exemplary real-life experiment, and then extended to a massive online dataset from Foursquare, proves our approach as feasible in estimating the tourist’s satisfaction with individual POIs. Moreover, it is already promising since incorporating location influence remains yet our future work and might further improve its performance.