This work analyzes about 35 million check-ins made by Foursquare users in over 5 million venues across the globe, and proposes a set of features that aim to capture the factors that may drive users' movements, finding that the supervised methodology based on the combination of multiple features offers the highest levels of prediction accuracy.
Abstract:
Mobile location-based services are thriving, providing an unprecedented opportunity to collect fine grained spatio-temporal data about the places users visit. This multi-dimensional source of data offers new possibilities to tackle established research problems on human mobility, but it also opens avenues for the development of novel mobile applications and services. In this work we study the problem of predicting the next venue a mobile user will visit, by exploring the predictive power offered by different facets of user behavior. We first analyze about 35 million check-ins made by about 1 million Foursquare users in over 5 million venues across the globe, spanning a period of five months. We then propose a set of features that aim to capture the factors that may drive users' movements. Our features exploit information on transitions between types of places, mobility flows between venues, and spatio-temporal characteristics of user check-in patterns. We further extend our study combining all individual features in two supervised learning models, based on linear regression and M5 model trees, resulting in a higher overall prediction accuracy. We find that the supervised methodology based on the combination of multiple features offers the highest levels of prediction accuracy: M5 model trees are able to rank in the top fifty venues one in two user check-ins, amongst thousands of candidate items in the prediction list.
TL;DR: A STAP model is proposed that first models the spatial and temporal activity preference separately, and then uses a principle way to combine them for preference inference, and a context-aware fusion framework is put forward to combine the temporal and spatial activity preference models for preferences inference.
TL;DR: This paper proposes social context mobile (SoCoMo) marketing as a new framework that enables marketers to increase value for all stakeholders at the destination to connect the different concepts of context-based marketing, social media and personalisation, as well as mobile devices.
TL;DR: A novel category-aware POI recommendation model is proposed, which exploits the transition patterns of users' preference over location categories to improve location recommendation accuracy and outperforms the state-of-the-artPOI recommendation models.
TL;DR: The WhoAmI method is proposed, a Double Dependent-Variable Factor Graph Model, to address the problem of double dependent-variable prediction-inferring user gender and age simultaneously, and shows that the proposed method significantly improves the prediction accuracy by up to 10% compared with several alternative methods.
TL;DR: The asymmetric impact of mobility and social relationships on predicting each other is discovered, which can serve as guidelines for future research on friendship and location prediction in LBSNs.
TL;DR: This work proposes a principled statistical framework for discerning and quantifying power-law behavior in empirical data by combining maximum-likelihood fitting methods with goodness-of-fit tests based on the Kolmogorov-Smirnov (KS) statistic and likelihood ratios.
TL;DR: This work identifies unique properties of implicit feedback datasets and proposes treating the data as indication of positive and negative preference associated with vastly varying confidence levels, which leads to a factor model which is especially tailored for implicit feedback recommenders.
TL;DR: A model of human mobility that combines periodic short range movements with travel due to the social network structure is developed and it is shown that this model reliably predicts the locations and dynamics of future human movement and gives an order of magnitude better performance.
TL;DR: A richer model relating geography and social-network friendship is introduced, in which the probability of befriending a particular person is inversely proportional to the number of closer people.
TL;DR: An on-line algorithm for learning preference functions that is based on Freund and Schapire's "Hedge" algorithm is considered, and it is shown that the problem of finding the ordering that agrees best with a learned preference function is NP-complete.
Q1. What are the contributions in "Mining user mobility features for next place prediction in location-based services" ?
This multi-dimensional source of data offers new possibilities to tackle established research problems on human mobility, but it also opens avenues for the development of novel mobile applications and services. In this work the authors study the problem of predicting the next venue a mobile user will visit, by exploring the predictive power offered by different facets of user behavior. The authors first analyze about 35 million check-ins made by about 1 million Foursquare users in over 5 million venues across the globe, spanning a period of five months. The authors then propose a set of features that aim to capture the factors that may drive users ’ movements. The authors further extend their study combining all individual features in two supervised learning models, based on linear regression and M5 model trees, resulting in a higher overall prediction accuracy. The authors find that the supervised methodology based on the combination of multiple features offers the highest levels of prediction accuracy: M5 model trees are able to rank in the top fifty venues one in two user check-ins, amongst thousands of candidate items in the prediction list.
Q2. What is the probability distribution of the intervals between consecutive check-ins?
Longer intervals are less likely than shorter ones, denoting that faster sequences of check-ins might arise, together with long periods of inactivity.
Q3. What is the average score of the features that exploit the spatial distance?
The Geographic Distance and Rank Distance attain an average score 0.78, highlighting that spatial distance is an important factor in the way users decide which venue to visit next.
Q4. How does the model tree perform in terms of accuracy?
Model trees excel in terms of prediction accuracy (shown here for N = 50, with all N shown in Figure 4), scoring above 0.5 in general, denoting that one in two user check-ins are successfully predicted.
Q5. What is the probability distribution of distance between check-ins?
The probability distribution of spatial distance between check-ins exhibits a decreasing trend (Figure 1(a)): shorter distances are more likely to appear.
Q6. What makes their approach suitable for prediction on new users with few or zero check-ins or?
This also makes their approach suitable for prediction on new users with few or zero check-ins or friends, thanks to features that make no use of historic user activity such as popularity and distance.
Q7. What is the method of training a model by providing feedback in the form of user preference?
This method of training a model by providing feedback in the form of user preference has been established in the past [4] and corresponds to an effective reduction of the ranking problem to a binary classification task.
Q8. how do the authors determine the rank score of a target venue?
the rank score of a target venue k is obtained by enumerating the past transitions observed by any user from the current location l′ to location k, which the authors formally define asr̂k(l ′) = |{(m,n) ∈ Lc : m = l′ ∧ n = k}| (8)Temporal Features.
Q9. What is the APR score for the linear regression model?
On the other hand, the linear regression model achieves an APR score equal to 0.81 which ranks it lower than the popularity and categorical preference features.