Performance Evaluation of LZ-Based Location Prediction Algorithms in Cellular Networks

“…There are many estimators for the entropy rate, but we will focus on the class of Lempel-Ziv (or LZ ) estimators, since they are known to rapidly converge to the true entropy rate and do not assume anything a priori about the statistics of the time series 2 [23,24]. The LZ estimator is closely related to LZ compression and prediction [25], however, it used for the arguably simpler task of quantifying the information present in a sequence.…”

Section: Information Theorymentioning

confidence: 99%

Breaking the habit: Measuring and predicting departures from routine in individual human mobility

McInerney

Stein

Rogers

et al. 2013

Pervasive and Mobile Computing

View full text Add to dashboard Cite

a b s t r a c tResearchers studying daily life mobility patterns have recently shown that humans are typically highly predictable in their movements. However, no existing work has examined the boundaries of this predictability, where human behaviour transitions temporarily from routine patterns to highly unpredictable states. To address this shortcoming, we tackle two interrelated challenges. First, we develop a novel information-theoretic metric, called instantaneous entropy, to analyse an individual's mobility patterns and identify temporary departures from routine. Second, to predict such departures in the future, we propose the first Bayesian framework that explicitly models breaks from routine, showing that it outperforms current state-of-the-art predictors.

show abstract

“…Rodriguez-Carrion [25,26] assessed three LZ-based algorithms by separating each algorithm into two independent phases (tree building and probability calculation) and further discussing hit rate and power consumption. Rodriguez-Carrion [25,26] assessed three LZ-based algorithms by separating each algorithm into two independent phases (tree building and probability calculation) and further discussing hit rate and power consumption.…”

Section: Introductionmentioning

confidence: 99%

“…In addition, the mobility prediction with low resource consumption is gaining more and more attentions. Rodriguez-Carrion [25,26] assessed three LZ-based algorithms by separating each algorithm into two independent phases (tree building and probability calculation) and further discussing hit rate and power consumption. Results demonstrated that LZ-based algorithms could learn mobility patterns and estimate the next place with low resource needs, which makes it possible to apply them on mobile devices.…”

Section: Introductionmentioning

confidence: 99%

User location prediction with energy efficiency model in the Long Term‐Evolution network

Qiao

Yang

et al. 2015

Int J Communication

View full text Add to dashboard Cite

Predicting users' next location/place allows us to anticipate their future movement. It provides additional time to be ready for that movement and react consequently. Furthermore, many industries, including Internet Service Providers, are still requiring low cost and simple location/place prediction methods that can be implemented on mobile device. This paper studies domain-independent prediction algorithms and spatiotemporal based prediction method using 20-day-long records in Long Term-Evolution(LTE) network, which captures the mobility patterns of 3474 individuals. After examining the prediction accuracy and resource consumption of domain-independent prediction algorithms, we find Markov provides the best tradeoff. Furthermore, Active LeZi outperforms Markov if enough consecutive parsed patterns of users' history movement are captured. In addition, we further group users according to their spatio-temporal entropy profiles in order to predict not only user's future locations but also the place he or she most likely to appear within a specific period. By applying the simple spatio-temporal based method to each group of user, 83.3% accuracy can be achieved for some users. Yet Markov and Active LeZi algorithms perform better for some other users. This implies that we should consider applying different prediction methods to users with distinct spatio-temporal characteristics. RELATED WORKIndividuals display significant regularity, as they tend to visit a few highly frequented locations, like home or office. Numerous studies show that people's movement trajectory is far from random. By measuring the entropy of each individual's trajectory, [1] achieved 93% potential predictability in user mobility. When considering both the frequencies and temporal correlations of individual movements, the theoretical maximum predictability can reach 88% [2].Previous studies regarding location prediction include several models and methods. The usage of well-known mobility models was originally applied in the area of location predication [3,4], like Bayesian approaches [5-7], neural networks [8], Hidden Markov models [9], Markov models [10] and compression algorithms [11-13]. In addition, some recently proposed new algorithms [14, 15] and frames [16-18] all presented very good results.Recently, regarding the location prediction, scholars start to consider many other factors that could influence the prediction results, like spatial context [4], temporal factors [19,20], spatio-temporal factors [21,22] and even demographics (such as gender and age) [23].

show abstract

Performance Evaluation of LZ-Based Location Prediction Algorithms in Cellular Networks

Cited by 21 publications

References 7 publications

FamTV: An architecture for presence-aware personalized television

FamTV: An architecture for presence-aware personalized television

Breaking the habit: Measuring and predicting departures from routine in individual human mobility

User location prediction with energy efficiency model in the Long Term‐Evolution network

Contact Info

Product

Resources

About