Models

CAPRI contains various models for analysis of the data in Point-of-Interest (POI) datasets. All of the models are available in the Models directory of the CAPRI package. Please note that As the framework is still in development, we are working on adding more models.

Baselines

TBD

Context-aware POI Recommendation

Context Aware Recommendation Systems incorporate a variety of contextual factors in order to accurately capture user preferences.

GeoSoCa

GoeSoCa is a novel POI recommendation method that uses geographical, social, and category correlations between users and POIs to make recommendations. These correlations can be learned from user check-in data on POIs in the past and used to predict a user’s relevance score to an unvisited POI in order to offer suggestions to them. Read more at GeoSoCa’s paper You can also check the content of GeoSoCa model in CAPRI package.

LORE

LORE is another model utilized in the context of context-aware POI recommendation systems. It is a popular and robust model for location recommendation focused on the impacts of geographical and social influence on users’ check-in behaviors. LORE incrementally mines sequential patterns from location sequences and represents the sequential patterns as a dynamic Location-Location Transition Graph (L2TG). It also predicts the probability of a user visiting a location by Additive Markov Chain (AMC) with L2TG. Finally, it fuses sequential influence with geographical influence and social influence into a unified recommendation framework. Read more at LORE’s paper You can also check the content of LORE model in CAPRI package.

USG

USG is a well-known model in the POI recommedner community. Due to the spatial clustering phenomenon demonstrated in LBSN user check-in activities, USG places a specific emphasis on geographical impact in addition to deriving user preference based on researching social influence from peers. Accordingly, geographical influence among POIs has a significant impact on user check-in behaviors, which is modedl in USG using a power law distribution. This model creates a naïve Bayesian-based collaborative recommendation system based on geographical influence. Read more at USG’s paper You can also check the content of USG model in CAPRI package.