MACACO: Mobile context-Adaptive CAching for COntent-centric networking

Lead Research Organisation: University College London
Department Name: Geography

Abstract

Finding new ways to manage the increased data usage and to improve the level of service required by the new wave of smartphones applications is an essential issue. The MACACO project proposes an innovative solution to this problem by focusing on data offloading mechanisms that take advantage of context and content information. Our intuition is that if it is possible to extract and forecast the behaviour of mobile network users in the three dimensional space of time, location and interest (i.e. 'what', 'when' and 'where' users are pulling data from the network), it is possible to derive efficient data offloading protocols. Such protocols would pre-fetch the identified data and cache them at the network edge at an earlier time, preferably when the mobile network is less congested, or offers better quality of service. Caching can be done directly at the mobile terminals, as well as at the edge nodes of the network (e.g., femtocells or wireless access points).

Building on previous research efforts in the fields of social wireless networking, opportunistic communications and content networking, MACACO will address several issues in this space. The first one is to derive appropriate models for the correlation between user interests and their mobility. Lots of studies have characterised mobile nodes mobility based on real world data traces, but knowledge about the interactions with user interests in this context is still missing. To fill this gap, MACACO proposes to acquire real world data sets to model mobile node behaviour in the aforementioned three-dimensional space. The second issue addressed is the derivation of efficient data-offloading algorithms leveraging the large-scale data traces and corresponding models. Firstly, simple and efficient prediction algorithms will be derived to forecast the node's mobility and interests. Then, MACACO will provide data pre-fetching mechanisms that both improves the perceived quality of service of the mobile user and
noticeably offloads peak bandwidth demands at the cellular network. A proof of concept will be exhibited though a federated testbed located in France, Switzerland and in the UK.

Planned Impact

The MACACO project is totally aligned with the expected types of impact of the second topic of the CHIST-ERA 2012 call. If successful, MACACO will foster new "services enabling the emergence of innovative network technologies" by providing the required context- and content-aware models and protocols to manage the increased data usage required by the new wave of smartphones applications. By doing so, MACACO aims to reinforce the European scientific excellence in the mobile service provision and helping European carriers to offload their cellular traffic.

Additionally, by detecting and modelling the correlations between user mobility and the traffic demand he/she generates, MACACO aims to strengthen the research field of mobile networks and human behaviour prediction and "develop a deeper fundamental and comprehensive understanding of new enhanced communication network architectures". Moreover, by facilitating development of new context- and content-aware applications, the project expects to foster significant innovation for industry. Novel and improved services will also have a significant for final users, given the fundamental role played by mobile Internet nowadays. In fact, several companies (from small start-ups to corporations) can benefit from improved wireless broadband services and from innovative content delivery mechanisms.

As also required by the call goals, MACACO "brings together researchers and research communities working on distinct network layers and on content and context extraction in the broader framework of a content- and context-adaptive communication networks". As stated before, the consortium was carefully constituted to gather partners that are pretty complementary and qualified to address the context-content correlation and related data offloading challenge. This constitutes one of the strengths of the project, which could not be conducted with the participation of only one or few of the involved partners. Thus, in addition to the technological impact, MACACO will have a significant impact in terms of competence building. The partners will combine research and experience in a wide set of areas to gain unique competence, which will be brought forward to other European partners through the dissemination and exploitation activities of the consortium.

Publications

10 25 50
publication icon
Baron B (2020) Where You Go Matters A Study on the Privacy Implications of Continuous Location Tracking in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies

publication icon
Cavdar S (2020) A Multi-perspective Analysis of Social Context and Personal Factors in Office Settings for the Design of an Effective Mobile Notification System in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies

publication icon
Darvariu V (2020) Quantifying the Relationships between Everyday Objects and Emotional States through Deep Learning Based Image Analysis Using Smartphones in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies

publication icon
Kandappu T (2020) PokeME

publication icon
Mehrotra A (2016) My Phone and Me

publication icon
Mehrotra A (2017) Interpretable Machine Learning for Mobile Notification Management An Overview of PrefMiner in GetMobile: Mobile Computing and Communications

publication icon
Mehrotra A (2021) FutureWare: Designing a Middleware for Anticipatory Mobile Computing in IEEE Transactions on Software Engineering

publication icon
Mehrotra A (2019) NotifyMeHere

 
Description MACACO is funded through the CHIST-ERA initiative and involves other partners namely INRIA, Toulouse, CNR and SUPSI Lugano.

Finding new ways to manage the increased data usage and to improve the level of service required by the new wave of smartphones applications is an essential issue. MACACO project (https://macaco.inria.fr/) proposes an innovative solution to this problem by focusing on data offloading mechanisms that take advantage of context and content information. Our intuition is that if it is possible to extract and forecast the behaviour of mobile network users in the three-dimensional space of time, location and interest (i.e. 'what', 'when' and 'where' users are pulling data from the network), it is possible to derive efficient data offloading protocols. Such protocols would pre-fetch the identified data and cache it at the network edge at an earlier time, preferably when the mobile network is less charged, or offers better quality of service. For that, we need to understand the relationship between human context (such as the his/her environment and actions) and their generated data traffic.

The preliminary results of these analysis have allowed us inferring the best times and places to transfer content from/to users' phones location and/or from/to the wireless infrastructure closest to the users' phones location. Other research activities were performed and consisted in the: energy and storage optimization of the application implementation; energy consumption analysis of different modules of the MACACO App; modelling, characterization and/or improvement (in terms of accuracy and temporal coverage) of cellular traces describing mobility and content demands of subscribers; spatiotemporal predictability of data traffic consumption of subscribers; profiling of users' interest in applications and web content dictating pre-fetching decisions at smartphone devices; profiling of users' context in terms of wireless connectivity dictating temporal pre-fetching decisions, i.e., when to bring content to smartphone devices; defining content transport strategies in urban environments using public transportation buses.

In particular, UCL developed a series of mechanisms for receiving the right information at the right time. In particular UCL focused on the study of mobile notifications and the analysis of user-interest. Indeed, Mobile notifications provide an effortless way to enable users to be aware of newly available information in quasi real- time. However, notifications arrive at inappropriate moments or carry irrelevant content. We analysed the impact of content on the acceptance of mobile notifications. The results of the analysis were used to develop a novel machine learning approach that predicts the acceptance of a notification by using its content and the context in which it is delivered. Such an approach can be used by an interruptibility management system to ensure that the right information is delivered at the right time. We collected notifications "in-the-wild" and classified them according to the information contained in them. Our results show that a user's activity can impact the time delay in the response to a notification. We evaluated the average percentage of the notifications that are accepted in a day for each notification category. Our results show that the chat notifications, where the sender is a family member or a relative of the user, have the highest acceptance rate. Our approach outperforms the user-defined rules for delivering notifications on their mobile phones. Finally, we implemented an online predictor in order to understand the required training period for successful prediction. We consider this work an initial step towards the implementation of an effective and efficient component for notification management for delivering the right information at the right time. This is a key mechanism for content-centric networking. We believe that the accuracy of the prediction model will increase by considering fine-grained categories that can be obtained by classifying the notification content using natural language processing techniques. This is particularly important for intelligent pre-fetching. Another orthogonal issue that will play a fundamental role in the development of these technology is privacy: indeed, the implementation of the proposed algorithms involve the analysis of user-generated content and of the social relationships of the users. Our goal is to explore ways for performing advanced processing on the phones in order to minimize the sharing of personal information with third-party entities.

We also worked on the automatic extraction of rules for intelligent notification management through an application called PrefMiner. The goal is to make notifications intelligible to users. We evaluated our proposed mechanism with a large-scale dataset of notifications collected during an interruptibility study. Our results show that by using the notification title and the user's location, we can predict if a message will be dismissed by a user with a very high precision. Through an in-the-wild deployment, we showed that PrefMiner represents a very effective, yet transparent, solution for interruptibility management for mobile devices. More recently, we have explored the role of places and context in the use of mobile phones, by means of large-scale study. More specifically, we analysed around 77000 location samples that were linked to 17000 instances of phone usage. Type of phone usage (calls, messaging, use of certain applications and so on) was used to cluster locations together. Finally, statistical tests were performed to examine whether certain types of place are linked to specific classes of interaction behavior. The results of this work could prove to be of key importance for the development of novel intelligent mobile applications and services and also for mobile application marketing purposes.

Finally, a paper about the design of intelligent notification systems in environments composed of multiple devices was presented at ACM CHIIR 2019 co-sponsored by SIGIR and SIGCHI. In 2020 we published a book ("Intelligent Notification Systems"), which also present results from this grant. The grant is acknowledged in the book.
Exploitation Route The findings of our work have a wide applicability in a variety of areas, from marketing to digital health (e.g., systems for positive behaviour intervention).

We released libraries for automatic notification delivery and learning of rules that are able to understand users' behavior from their application usage and their context. This work is currently under way.

The work has been (and will be) widely disseminated through conferences (such as UbiComp, the leading international conference in Ubiquitous Computing). We recently published a book for Morgan&Claypool on intelligent notifications, which is the result of the work from this grant. We believe that this is a great resources for other researchers and practitioners interested in applying the findings of this project.

We won two Best Paper Awards for work associated to this project:

- Best Paper Award at UbiComp 2016 for our paper "PrefMiner: Mining User's Preferences for Intelligent Mobile Notification Management". The paper has has been listed in the ACM Best of Computing Notable Books and Articles for 2016.

- Best Paper Award at ACM CHIIR'19 for the paper "NotifyMeHere: Intelligent Notification Delivery in Multi-Device Environments".
Sectors Digital/Communication/Information Technologies (including Software),Leisure Activities, including Sports, Recreation and Tourism

URL http://macaco.inria.fr
 
Description We realised our codebase and applications we have developed a series of prototypes related to this project: - QuickLaunch; - PrefMiner; - MyLifeLogger; - NotifyMe. The code of our software is available on the GitHub site of the Intelligent Social Systems Lab (the lab led by the PI Mirco Musolesi): https://github.com/nsds Even if there was not a direct exploitation of the findings, some of the techniques used in this project has been used as a concept for applications like for example Moody (https://moodymonth.com). Mirco Musolesi collaborated to the design of the sensing part of the app. We published a book ("Intelligent Notification Systems", published by Morgan&Claypool), which also targets practitioners. The book contains a description of many findings of MACACO, which can be applied to a variety of real world applications.
First Year Of Impact 2020
Sector Creative Economy,Digital/Communication/Information Technologies (including Software),Leisure Activities, including Sports, Recreation and Tourism
Impact Types Societal,Economic

 
Description Royal Society International Travel Funding
Amount £12,000 (GBP)
Organisation The Royal Society 
Sector Charity/Non Profit
Country United Kingdom
Start 08/2016 
End 08/2018
 
Title My Phone and Me 
Description My Phone and Me can help you to monitor your phone usage pattern and your interaction with notifications. The app presents an avatar which tells your addiction level at the current moment. Also, it shows the amount of time the phone has been used, the app which is used the most and the app which triggers most notifications. Moreover, the app offers you to visualise your phone activities such as hourly phone usage, hourly usage of an individual app, and your interaction with notifications. 
Type Of Technology Webtool/Application 
Year Produced 2015 
Impact The application allows users to monitor their interactions with mobile phones in order to better understand the role these devices play in their daily life and act accordingly. More in general, thanks to the data collected by means of this application we were able to understand the behaviour of mobile users in a very fine-grained detail. 
URL https://play.google.com/store/apps/details?id=com.nsds.myphoneandme
 
Title NotifyMe 
Description NotifyMe is an application for statistical analysis of mobile notification receptivity. It aims to collect data about user responses to mobile notifications and the context of notification delivery in order to study the correlation between notification's content category, context of delivery, and the notification receptivity. 
Type Of Technology Webtool/Application 
Year Produced 2015 
Impact This is an application that allowed us to collect information about user interaction with notifications. This knowledge can be exploited in order to improve our interactions with notifications. 
URL https://play.google.com/store/apps/details?id=com.nsds.notifyme&hl=en
 
Title QuickLaunch 
Description Quick Launch is a predictive lock screen for a quick access of apps. Key functionalities: - Presents application icons on the lock screen for quick access; - Dynamically arrange apps based on your needs during the day. 
Type Of Technology Webtool/Application 
Year Produced 2016 
Impact The application allowed us to test predictive testing in order to develop anticipatory mechanisms for mobile apps. The application has also a practical use, since it allows users to pre-load applications on the lock screen anticipating their needs given a certain context. 
URL https://play.google.com/store/apps/details?id=com.quick.lauch&hl=en