Across different web services, we hear about recommendation systems that help users tackle information overload. Travel is different: the world does not have millions of cities, but finding new, interesting places to inspire people to travel is still a challenge. What are the factors that make traveling to destinations appealing, and how do those factors change based on your origin? What data, algorithms, and interactions do we need to surface destinations as recommendations? Moreover, how can a recommender system be built in a domain where typical users will book a flight, anonymously, less than a handful of times per year?

Measuring mental well-being with mobile sensing has been an increasingly active research topic. Pervasiveness of smartphones combined with the convenience of mobile app distribution platforms (e.g., Google Play) provide a tremendous opportunity to reach out to millions of users. However, the studies at the confluence of mental health and mobile sensing have been longitudinally limited, controlled, or confined to a small number of participants. In this paper we report on what we believe is the largest longitudinal in-the-wild study of mood through smartphones. We describe an Android app to collect participants' self-reported moods and system triggered experience sampling data while passively measuring their physical activity, sociability, and mobility via their device's sensors. We report the results of a large-scale analysis of the data collected for about three years from 18,000 users.

Public transport information systems have been shown to positively affect passengers' usage of their city's transport infrastructure, by providing information such as the location and schedules of trains and buses. These systems, however, lack qualitative information about passengers' ongoing experi

The Experience Sampling Method (ESM) has been widely used to collect longitudinal survey data from participants; in this domain, smartphone sensors are now used to augment the context-awareness of sampling strategies. In this paper, we study the effect of ESM design choices on the inferences that can be made from participants' sensor data, and on the variance in survey responses that can be collected. In particular, we answer the question: are the behavioural inferences that a researcher makes with a trigger-defined subsample of sensor data biased by the sampling strategy's design? We demonstrate that different single-sensor sampling strategies will result in what we refer to as contextual dissonance: a disagreement in how much different behaviours are represented in the aggregated sensor data. These results are not only relevant to researchers who use the ESM, but call for future work into strategies that may alleviate the biases that we measure.

Increasingly, web recommender systems face scenarios where they need to serve suggestions to groups of users; for example, when families share e-commerce or movie rental web accounts. Research to date in this domain has proposed two approaches: computing recommendations for the group by merging any members' ratings into a single profile, or computing ranked recommendations for each individual that are then merged via a range of heuristics. In doing so, none of the past approaches reason on the preferences that arise in individuals when they are members of a group. In this work, we present a probabilistic framework, based on the notion of information matching, for group recommendation. This model defines group relevance as a combination of the item's relevance to each user as an individual and as a member of the group; it can then seamlessly incorporate any group recommendation strategy in order to rank items for a set of individuals. We evaluate the model's efficacy at generating recommendations for both single individuals and groups using the MovieLens and MoviePilot data sets. In both cases, we compare our results with baselines and state-of-the-art collaborative filtering algorithms, and show that the model outperforms all others over a variety of ranking metrics.

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.

Mobile, local recommendations are on the rise. Surprisingly however, research addressing user perceptions of local recommendations and local differences when interacting with such recommendation services is yet scarce. Location-based recommendation services are mostly evaluated from a `recommendation systems' standpoint, with limited experiential insights from users and limited focus on local differences that may apply. This workshop focuses on the local, personal user experience, and provides a forum to exchange experiences, insights and strategies in personalizing local mobile applications and generating local recommendations that fit local user needs.

The widespread adoption of mobile technology allows personalised applications to be deployed in an increasing host of contexts; user modelling, profiling, and personalised recommendations are becoming an integral component of mobile information systems. Furthermore, mobile technology enables the recording and collection of facets of daily life, which has given rise to the notion of the quantified self; researchers operating at the intersection of computer and social science are now seeking to understand how these mobiles' data can aide the design of health interventions and inform future psychological and social science research. In this work, we describe the design of a personalised mobile application that seeks user feedback and builds a user profile about people's gastrointestinal health using ratings and tags. We describe the application's design and the personalised health insights it provides (and, particularly, why recommendations were not designed as a means for self-diagnosis).

The aim of a web-based recommender system is to provide highly accurate and up-to-date recommendations to its users; in practice, it will hope to retain its users over time. However, this raises unique challenges. To achieve complex goals such as keeping the recommender model up-to-date over time, we need to consider a number of external requirements. Generally, these requirements arise from the physical nature of the system, for instance the available computational resources. Ideally, we would like to design a system that does not deviate from the required outcome. Modeling such a system over time requires to describe the internal dynamics as a combination of the underlying recommender model and the its users' behavior. We propose to solve this problem by applying the principles of modern control theory - a powerful set of tools to deal with dynamical systems - to construct and maintain a stable and robust recommender system for dynamically evolving environments. In particular, we introduce a design principle by focusing on the dynamic relationship between the recommender system's performance and the number of new training samples the system requires. This enables us to automate the control other external factors such as the system's update frequency. We show that, by using a Proportional-Integral-Derivative controller, a recommender system is able to automatically and accurately estimate the required input to keep the output close to a pre-defined requirements. Our experiments on a standard rating dataset show that, by using a feedback loop between system performance and training, the trade-off between the effectiveness and efficiency of the system can be well maintained. We close by discussing the widespread applicability of our approach to a variety of scenarios that recommender systems face.

The widespread adoption of automated fare collection (AFC) systems by public transport authorities around the world means that, increasingly, people carry and use passive sensors (embedded inside of public transit tickets) to record their daily movements. Unlike mobile phones, the records held by AFC systems provide a rich and detailed source of data about peoples' transport habits: times of travel, modalities, destinations, trip durations, and fares paid. In this work, we explore the extent that this data offers the possibility to both build and measure future of travel-based ubiquitous computing applications. We focus on two potential end-users: first, how travellers may be aided by feedback mechanisms in order to re-align misperceptions of their travel behaviour and leverage this data to change their habits. In particular, we analyse differences between 85 travellers' surveyed perceptions of their public transport habits and their actual usage of the system. Second, how transport authorities can use this data to measure and implement incentive mechanisms that produce the expected impact. We use anonymised AFC data to measure the extent that financial incentives implemented by London's transport authority (such as peak-hour fares and student discounts) correlate with measurable changes in millions of travellers' behaviours.

As the public transport infrastructure of large cities expands, transport operators are diversifying the range and prices of tickets that can be purchased for travel. However, selecting the best fare for each individual traveller's needs is a complex process that is left almost completely unaided. By examining the relation between urban mobility and fare purchasing habits in large datasets from London, England's public transport network, we estimate that travellers in the city cumulatively spend, per year, up to approximately GBP 200 million more than they need to, as a result of purchasing the incorrect fares. We propose to address these incorrect purchases by leveraging the huge volumes of data that travellers create as they move about the city, by providing, to each of them, personalised ticket recommendations based on their estimated future travel patterns. In this work, we explore the viability of building a fare-recommendation system for public transport networks by (a) formalising the problem as two separate prediction problems and (b) evaluating a number of algorithms that aim to match travellers to the best fare. We find that applying data mining techniques to public transport data has the potential to provide travellers with substantial savings.

The rise of location-enabled mobile phones and location based services offers a great opportunity to apply personalization and recommender system technology to people's everyday lives. A variety of digital traces can now be used to infer how people move about their city and extract their context and habits. Personalization and recommender systems, potentially merged with the data that people store online (e.g., social networks, web ratings), can then not only be used to recommend new places and events that they may find interesting to attend, but, more broadly, personalize and enhance any service that people find themselves using.

Traveller information, route planning, and service updates have become essential components of public transport systems: they help people navigate built environments by providing access to information regarding delays and service disruptions. However, one aspect that these systems lack is a way of tailoring the information they offer in order to provide personalised trip time estimates and relevant notifications to each traveller. Mining each user's travel history, collected by automated ticketing systems, has the potential to address this gap. In this work, we analyse one such dataset of travel history on the London underground. We then propose and evaluate methods to (a) predict personalised trip times for the system users and (b) rank stations based on future mobility patterns, in order to identify the subset of stations that are of greatest interest to the user and thus provide useful travel updates.

A city offers thousands of social events a day, and it is difficult for dwellers to make choices. The combination of mobile phones and recommender systems can change the way one deals with such abundance. Mobile phones with positioning technology are now widely available, making it easy for people to broadcast their whereabouts, recommender systems can now identify patterns in people's movements in order to, for example, recommend events. To do so, the system relies on having mobile users who share their attendance at a large number of social events: cold-start users, who have no location history, cannot receive recommendations. We set out to address the mobile cold-start problem by answering the following research question: how can social events be recommended to a cold-start user based only on his home location? To answer this question, we carry out a study of the relationship between preferences for social events and geography, the first of its kind in a large metropolitan area. We sample location estimations of one million mobile phone users in Greater Boston, combine the sample with social events in the same area, and infer the social events attended by 2,519 residents. Upon this data, we test a variety of algorithms for recommending social events. We find that the most effective algorithm recommends events that are popular among residents of an area. The least effective, instead, recommends events that are geographically close to the area. This last result has interesting implications for location-based services that emphasize recommending nearby events.

Collaborative Filtering (CF) algorithms, used to build web-based recommender systems, are often evaluated in terms of how accurately they predict user ratings. However, current evaluation techniques disregard the fact that users continue to rate items over time: the temporal characteristics of the system's top-N recommendations are not investigated. In particular, there is no means of measuring the extent that the same items are being recommended to users over and over again. In this work, we show that temporal diversity is an important facet of recommender systems, by showing how CF data changes over time and performing a user survey. We then evaluate three CF algorithms from the point of view of the diversity in the sequence of recommendation lists they produce over time. We examine how a number of characteristics of user rating patterns (including profile size and time between rating) affect diversity. We then propose and evaluate set methods that maximise temporal recommendation diversity without extensively penalising accuracy.

Nearest-neighbor collaborative filtering provides a successful means of generating recommendations for web users. However, this approach suffers from several shortcomings, including data sparsity and noise, the cold-start problem, and scalability. In this work, we present a novel method for recommending items to users based on expert opinions. Our method is a variation of traditional collaborative filtering: rather than applying a nearest neighbor algorithm to the user-rating data, predictions are computed using a set of expert neighbors from an independent dataset, whose opinions are weighted according to their similarity to the user. This method promises to address some of the weaknesses in traditional collaborative filtering, while maintaining comparable accuracy. We validate our approach by predicting a subset of the Netflix data set. We use ratings crawled from a web portal of expert reviews, measuring results both in terms of prediction accuracy and recommendation list precision. Finally, we explore the ability of our method to generate useful recommendations, by reporting the results of a user-study where users prefer the recommendations generated by our approach.

Collaborative Filtering aims to predict user tastes, by minimising the mean error produced when predicting hidden user ratings. The aim of a deployed recommender system is to iteratively predict users' preferences over a dynamic, growing dataset, and system administrators are confronted with the problem of having to continuously tune the parameters calibrating their CF algorithm. In this work, we formalise CF as a time-dependent, iterative prediction problem. We then perform a temporal analysis of the Netflix dataset, and evaluate the temporal performance of two CF algorithms. We show that, due to the dynamic nature of the data, certain prediction methods that improve prediction accuracy on the Netflix probe set do not show similar improvements over a set of iterative train-test experiments with growing data. We then address the problem of parameter selection and update, and propose a method to automatically assign and update per-user neighbourhood sizes that (on the temporal scale) outperforms setting global parameters.

Recommender systems, based on collaborative filtering, draw their strength from techniques that manipulate a set of user-rating profiles in order to compute predicted ratings of unrated items. There are a wide range of techniques that can be applied to this problem; however, the k-nearest neighbour (kNN) algorithm has become the dominant method used in this context. Much research to date has focused on improving the performance of this algorithm, without considering the properties that emerge from manipulating the user data in this way. In order to understand the effect of kNN on a user-rating dataset, the algorithm can be viewed as a process that generates a graph, where nodes are users and edges connect similar users: the algorithm generates an implicit social network amongst the system subscribers. Temporal updates of the recommender system will impose changes on the graph. In this work we analyse user-user kNN graphs from a temporal perspective, retrieving characteristics such as dataset growth, the evolution of similarity between pairs of users, the volatility of user neighbourhoods over time, and emergent properties of the entire graph as the algorithm parameters change. These insights explain why certain kNN parameters and similarity measures outperform others, and show that there is a surprising degree of structural similarity between these graphs and explicit user social networks.

Collaborative filtering has become an established method to measure users' similarity and to make predictions about their interests. However, prediction accuracy comes at the cost of user's privacy: in order to derive accurate similarity measures, users are required to share their rating history with each other. In this work we propose a new measure of similarity, which achieves comparable prediction accuracy to the Pearson correlation coefficient, and that can successfully be estimated without breaking users' privacy. This novel method works by estimating the number of concordant, discordant and tied pairs of ratings between two users with respect to a shared random set of ratings. In doing so, neither the items rated nor the ratings themselves are disclosed, thus achieving strictly-private collaborative filtering. The technique has been evaluated using the recently released Netflix prize dataset.