Context-prediction performance by a dynamic Bayesian network: Emphasis on location prediction in ubiquitous decision support environment☆
Introduction
As the ubiquity becomes perceived by users more easily in their daily lives, context-aware systems have received a lot of attention from both practitioners and researchers. World-wild excitement about smart phones is also representing people’s passion about the context-aware devices. In this respect, context-awareness has been the subject of the growing attention in the area of ubiquitous computing over the years due to its usefulness for several application domains (Hong, Suh, & Kim, 2008). When computer systems are aware of the context in which they are used and are able to adapt to changes in context, they can engage in more efficient interaction with users.
Context awareness is concerned with enabling ubiquitous computing devices to be aware of changes in the environment, and to intelligently adapt themselves to provide more meaningful and timely decision support for decision-makers (Feng, Teng, & Tan, 2009). However, context-aware systems are limited by the fact that their target is the current context, and that the future context is not predicted by context-aware systems. Therefore, the quality of services provided by the context-aware systems is seriously restricted when future contexts change drastically. To this end, we need to consider the task of context prediction in order to proactively offer high-quality services for users in ubiquitous computing environments.
Context prediction opens a wide variety of possibilities of context-aware computing applications. A context-prediction application may infer the future location of an office owner and redirect incoming calls to the future location. A context-prediction application may also be useful for enhancing the quality of transportation systems. Based on the information about the current location and the future location of a particular user, transportation systems equipped with context prediction technology may be able to assist drivers more effectively by inferring possible preferred routes and by providing customized route suggestions for drivers, as well as warning the drivers about possible dangers by predicting their future context. Knowing the current location and current time, together with the user’s calendar, could also allow application to have a good idea of the user’s current social situation, such as if the user is in a meeting, in class, waiting in the airport, and so on.
The list of applications listed here is limited and we believe that there is a great potential for context prediction to be used in a variety of ubiquitous computing applications. Especially, it becomes clear how much users would benefit from the ubiquitous decision support systems equipped with the context-prediction mechanism. As an alternative for the inference engine to be used in the ubiquitous decision support systems capable of providing context-prediction function, this paper proposes a dynamic Bayesian network (DBN) approach to location prediction for ubiquitous computing environments. DBN is an important technique because of its ability to represent the temporal properties of user context information. In fact, it is obvious that a user’s current locations are influenced by their previous locations, and particular locations afford particular types of actions. Therefore, we adopted a DBN approach for recognizing the locations of users.
This paper is structured as follows. Section 2 discusses context prediction and various context prediction techniques in ubiquitous computing environments. The modeling techniques used to predict a user’ locations are described in Section 3. The results of the experiment are presented and discussed in Section 4, followed by concluding remarks and directions for future work in Section 5.
Section snippets
Context prediction
Context prediction focuses on inferring users’ context based on analyzing the observed context history that users have shown so far. The observed context history is a series of context information showing how users are moving around in a certain ubiquitous computing environment. The context information is supplied by various types of sensors such as GPS, RFID, and a variety of wireless devices. These sensors may provide the context information about users’ locations, users’ actions, or the
Inducing location prediction models
Many types of location recognition models can be learned. We investigated probabilistic models such as dynamic Bayesian networks (DBNs), general Bayesian networks (GBNs), tree augmented Naïve Bayesian networks (TANs), and Naïve Bayesian networks (NBNs). Refer to appendix for more details about Bayesian networks that were considered in this study.
Evaluation
In a formal evaluation, data were gathered from 336 subjects (undergraduate students at a private university in Seoul, Korea). In order to fully engage participation of subjects, two percent of the subject’s total class points were given as extra credit points. There were 125 female and 211 male participants of varying ages. The average age of female subjects was 20.7 years old, while the average age of the male subjects was 22.6 years old.
After filling out a demographic survey, participants were
Conclusion and Future work
Context prediction is an important problem in ubiquitous computing environments. Accurately predicting user contexts could greatly improve the quality of user satisfaction in every aspect of daily life, particularly in the use of ubiquitous decision support systems. By drawing inferences about user locations, ubiquitous decision support systems should not only automatically detect a user’s current situation, but also forecast the user’s likely future location. Such location prediction systems
References (26)
- et al.
Modeling situation awareness for context-aware decision support
Expert Systems with Applications
(2009) - et al.
Landmark detection from mobile life log using a modular Bayesian network model
Expert Systems with Applications
(2009) A neural network approach to movement pattern analysis
Human Movement Science
(2004)- Anagnostopoulos, T., Anagnostopoulos, C., Hadjiefthymiades, S., Kyriakakos, M., & Kalousis, A. (2009). Predicting the...
- Brdiczka, O., Reignier, P., & Crowley, J. (2007). Detecting individual activities from video in a smart home. In 11th...
- et al.
Approximating discrete probability distributions with dependence trees
IEEE Transactions on Information Theory
(1968) - et al.
A Bayesian method for the induction of probabilistic networks from data
Machine Learning
(1992) - Davison, B. D., & Hirsh, H. (1998). Predicting sequences of user actions. In AAAI/ICML workshop on predicting the...
- Fung, R., & Crawford, S. (1990). Constructor: A system for the induction of probabilistic models. In Proceedings of the...
- Heckerman, D. (1995). A Bayesian approach to learning causal networks. In Proceedings of the 11th conference on...
Context-aware systems: A literature review and classification
Expert Systems with Applications
Cited by (18)
An anomalous sound detection methodology for predictive maintenance
2022, Expert Systems with ApplicationsCitation Excerpt :Future works will be devoted to analyze different types of conditioning networks (i.e. Variational AutoEncoders (VAE) or Generative Adversarial Networks (GANs)) and operations together with the application of several pre-processing strategies to get better training performances, like noise reduction, with the aim of reducing the audio clips background noise surely present in factory environments, or audio data augmentation techniques, as well as pitching, time-shifting and so on. Furthermore, we will explore two emergent research topic called Context Prediction and Context Histories which are based on time series of Contexts that can be used to record in context histories the data of machines and use this data for different kinds of data analysis (Filippetto, Lima, & Barbosa, 2021; Lee & Lee, 2012; da Rosa, Barbosa, & Ribeiro, 2016). Emanuele Di Fiore: Conceptualization, Methodology, Software, Validation, Formal analysis, Data curation, Writing – original draft, Writing - review & editing.
ORACON: An adaptive model for context prediction
2016, Expert Systems with ApplicationsCitation Excerpt :Expert and intelligent systems have applied prediction to specific problems, such as trade of stocks (Ballings et al., 2015; Rather et al., 2015; Wang et al., 2015), software failure treatment (Bala & Chana, 2015), and prediction in housing prices (Park & Bae, 2015). In addition, time series (Al-Hmouz et al., 2015; Ma et al., 2015; Wu & Lee, 2015) and location prediction have been receiving attention from researchers (Burbey & Martin, 2012; David et al., 2013; Lee & Lee, 2012). Recent studies discussed research trends in the context prediction (Ameyed et al., 2015; Pejovic & Musolesi, 2015; VanSyckel & Becker, 2014).
A decision support model to evaluate liveability in the context of urban vibrancy
2022, International Journal of Architectural ComputingModular and Personalized Smart Health Application Design in a Smart City Environment
2018, IEEE Internet of Things JournalPredictive ridesharing based on personal mobility patterns
2017, IEEE Intelligent Vehicles Symposium, ProceedingsMobiDict - A Mobility prediction system leveraging realtime location data streams
2016, Proceedings of the 7th ACM SIGSPATIAL International Workshop on GeoStreaming, IWGS 2016
- ☆
This research is supported by the Ubiquitous Computing and Network (UCN) Project, Knowledge and Economy Frontier R&D Program of the Ministry of Knowledge Economy (MKE) in Korea as a result of UCN’s subproject 11C3-T2-20S.