Personalized location models with adaptive mixtures

Abstract

Personalization is increasingly important for a range of applications that rely on location-based modeling. A key aspect in building personalized models is using populationlevel information to smooth noisy sparse data at the individual level. In this paper we develop a general mixture model framework for learning individual-level location models where the model adaptively combines different types of smoothing information. In a series of experiments with Twitter geolocation data and Gowalla check-in data we demonstrate that the proposed approach can be significantly more accurate than more traditional smoothing and matrix factorization techniques. The improvement in performance over matrix factorization is pronounced and may be explained by the tendency of dimensionality reduction methods to oversmooth and not retain enough detail at the individual level. CCS Concepts •Computing methodologies → Mixture models;

A Hybrid Generative Model for Online User Behavior Prediction

“…Inspired by this research, repeat consumption analysis has been applied to many other domains, including music listening [31], video watching [32], location visiting [33].…”

“…The personal consumption history component accounts for predicting repeat consumption, while the other component affects the prediction of novel consumption. Personalized Location Models [33] is another mixture model that combines the location history of users and additional information such as population pattern, geographic constraints, and social context to predict the future location of users.…”

A Variational Autoencoder Mixture Model for Online Behavior Recommendation

“…In [5], the authors developed a multinomial mixture model of history consumption of user and popularity of items to predict future consumption with different types of datasets, including location, music, or topic. The authors in [6] combined the locations history, the population pattern, the geographical constraint, and other social contexts to predict future locations.…”

“…When applying mixture models to behavior recommendation, we assume that the behaviors of users are affected by multiple factors with different impact levels, for example, the users' habits or their interests. Multinomial mixture model [5], Adaptive mixtures model [6], Hybrid generative model [7] are some recent works that investigate this problem by proposing a mixture model consisting of the user's behavior history, the popularity of behavior or geographic distance as their components.…”