Prediction Models to Enhance Location Based Services in Urban Areas

Karimzadeh, Mostafa (2020). Prediction Models to Enhance Location Based Services in Urban Areas. (Dissertation, Institute for computer science, Faculty of Science)

Text thesis_final.pdf - Other Restricted to registered users only Available under License BORIS Standard License. Download (14MB)

Mobility trace data typically includes historical information of the user's visited locations, which presents the visited places' detailed context and corresponding time-stamps. This collected data of moving objects (e.g, pedestrians, vehicles) enables new opportunities for prediction models to capture and estimate future movement patterns of individual or group users. Predicting future behaviour of moving objects can foster Location-based Services (LBSs). LBSs can be defined as services that use the location of users to bring location-specific and personalized services and information to them.

To have successful LBSs, first, we need to extract frequently visited locations of users (e.g., home, workplace, etc.) from accumulated mobility traces. This source of information provides possibilities for building models to learn and estimate the future mobility of an individual or group of moving objects in large cities. However, due to spatio-temporal dependencies of urban environments and the time-varying characteristics of user’s movement patterns, it is still challenging to discover urban hotspots (frequently visited locations), and predicting moving object’s future movement.

This thesis aims to define prediction algorithms to accurately estimate future behaviour for moving objects in urban areas. We benefit from this information to improve LBSs. This thesis key contributions can be summarized as follows: The first contribution is to detect users’ hotspots in urban areas. We utilize spatio-temporal analysis on collected geo-location points to discover moving objects’ frequently visited areas in city environments. The second contribution is to estimate the future location of pedestrians in the city environment. In this way, we design a mobility prediction algorithm that benefits from both first and second-order Markov chain algorithms. Third, we investigate predictors to estimate the future trajectory of moving objects (e.g., vehicles, pedestrians) in urban areas. A trajectory is a path that a moving object takes to travel from one location to another one. To make a trajectory prediction, we proposed two novel algorithms that are based on Markov chain and neural network algorithms. Fourth, we explore algorithms to estimate traffic flow on urban trajectories. A traffic flow estimator attempts to predict the future state of urban traffic in terms of the number of users in the trajectories. We propose two algorithms that can efficiently estimate future states of urban traffic. Additionally, we introduce two system models that benefit from mobility predictor and traffic flow predictor to enhance content prefetching and safety data dissemination for users, respectively. We evaluate our proposed algorithms and system models using two real-world and large-scale datasets collected from mobile networks and VANETs.

Interest & Impact

Downloads

31 since deposited on 11 Dec 2020
4 in the past 12 months

Citations

5 Citations in Web of Science ®
6 Citations in Scopus

Search

in Google Scholar™

Services

Actions (login required)

Edit item

Actions (login required)

Edit item