Context-aware mobility analytics and trip planning

Publication Year: 2018 Publication Type : Thesis

Abstract:


The study of user mobility is to understand and analyse the movement of individuals in the spatial and temporal domains. Mobility analytics and trip planning are two vital components of user mobility that facilitate the end users with easy to access navigational support through the urban spaces and beyond. Mobility context describes the situational factors that can influence user mobility decisions. The context-awareness in mobility analytics and trip planning enables a wide range of end users to make effective mobility decisions. With the ubiquity of urban sensing technologies, various situational factors related to user mobility decisions can now be collected at low cost and effort. This huge volume of data collected from heterogeneous data sources can facilitate context-aware mobility analytics and trip planning through intelligent analysis of mobility contexts, mobility context prediction, mobility context representation and integration considering different user perspectives. In each chapter of this thesis such issues are addressed through the development of case-specific solutions and real-world deployments.

Mobility analytics include prediction and analysis of many diverse mobility contexts. In this thesis, we present several real-world user mobility scenarios to conduct intelligent contextual analysis leveraging existing statistical methods. The factors related to user mobility decisions are collected and fused from various publicly available open datasets. We also provide future prediction of important mobility contexts which can be utilized for mobility decision making. The performance of context prediction tasks can be affected by the imbalance in context distribution. Another aspect of context prediction is that the knowledge from domain experts can enhance the prediction performance however, it is very difficult to infer and incorporate into mobility analytics applications. We present a number of data-driven solutions aiming to address the imbalanced context distribution and domain knowledge incorporation problems for mobility context prediction. Given an imbalanced dataset, we design and implement a framework for context prediction leveraging existing data mining and sampling techniques. Furthermore, we propose a technique for incorporating domain knowledge in feature weight computation to enhance the task of mobility context prediction.


In this thesis, we address key issues related to trip planning. Mobility context inference is a challenging problem in many real-world trip planning scenarios. We introduce a framework that can fuse contextual information captured from heterogeneous data sources to infer mobility contexts. In this work, we utilize public datasets to infer mobility contexts and compute trip plans. We propose graph based context representation and query based adaptation techniques on top of the existing methods to facilitate trip planning tasks. The effectiveness of trip plans relies on the efficient integration of mobility contexts considering different user perspectives. Given a contextual graph, we introduce a framework that can handle multiple user perspectives concurrently to compute and recommend trip plans to the end user.


This thesis contains efficient techniques that can be employed in the area of urban mobility especially, context-aware mobility analytics and trip planning. This research is built on top of the existing predictive analytics and trip planning techniques to solve problems of contextual analysis, prediction, context representation and integration in trip planning for real-world scenarios. The contributions of this research enable data-driven decision support for traveling smarter through urban spaces and beyond.


BibTex:

@phdthesis{SaiedurRahamanMohammad2018Cmaa, title = {Context-aware mobility analytics and trip planning},
    abstract = {The study of user mobility is to understand and analyse the movement of individuals in the spatial and temporal domains. Mobility analytics and trip planning are two vital components of user mobility that facilitate the end users with easy to access navigational support through the urban spaces and beyond. Mobility context describes the situational factors that can influence user mobility decisions. The context-awareness in mobility analytics and trip planning enables a wide range of end users to make effective mobility decisions. With the ubiquity of urban sensing technologies, various situational factors related to user mobility decisions can now be collected at low cost and effort. This huge volume of data collected from heterogeneous data sources can facilitate context-aware mobility analytics and trip planning through intelligent analysis of mobility contexts, mobility context prediction, mobility context representation and integration considering different user perspectives. In each chapter of this thesis such issues are addressed through the development of case-specific solutions and real-world deployments.

Mobility analytics include prediction and analysis of many diverse mobility contexts. In this thesis, we present several real-world user mobility scenarios to conduct intelligent contextual analysis leveraging existing statistical methods. The factors related to user mobility decisions are collected and fused from various publicly available open datasets. We also provide future prediction of important mobility contexts which can be utilized for mobility decision making. The performance of context prediction tasks can be affected by the imbalance in context distribution. Another aspect of context prediction is that the knowledge from domain experts can enhance the prediction performance however, it is very difficult to infer and incorporate into mobility analytics applications. We present a number of data-driven solutions aiming to address the imbalanced context distribution and domain knowledge incorporation problems for mobility context prediction. Given an imbalanced dataset, we design and implement a framework for context prediction leveraging existing data mining and sampling techniques. Furthermore, we propose a technique for incorporating domain knowledge in feature weight computation to enhance the task of mobility context prediction.


In this thesis, we address key issues related to trip planning. Mobility context inference is a challenging problem in many real-world trip planning scenarios. We introduce a framework that can fuse contextual information captured from heterogeneous data sources to infer mobility contexts. In this work, we utilize public datasets to infer mobility contexts and compute trip plans. We propose graph based context representation and query based adaptation techniques on top of the existing methods to facilitate trip planning tasks. The effectiveness of trip plans relies on the efficient integration of mobility contexts considering different user perspectives. Given a contextual graph, we introduce a framework that can handle multiple user perspectives concurrently to compute and recommend trip plans to the end user.


This thesis contains efficient techniques that can be employed in the area of urban mobility especially, context-aware mobility analytics and trip planning. This research is built on top of the existing predictive analytics and trip planning techniques to solve problems of contextual analysis, prediction, context representation and integration in trip planning for real-world scenarios. The contributions of this research enable data-driven decision support for traveling smarter through urban spaces and beyond.},
   
    author = {Saiedur Rahaman, Mohammad},
    keywords = {Context-aware Mobility;Accessible trip planning;Wait time prediction;Queue context prediction;Heterogeneous data fusion;Machine learning;Predictive analytics},
    language = {eng},
    school = {RMIT University},
    year = {2018},
   
}

Cite:

Related Publications

RUP: Large Room Utilisation Prediction with carbon dioxide sensor
Type : JournalArticle
Show More
A Scalable Room Occupancy Prediction with Transferable Time Series Decomposition of CO 2 Sensor Data
Type : JournalArticle
Show More
Topical Event Detection on Twitter
Type : ConferenceProceeding
Show More

© 2021 Flora Salim - CRUISE Research Group.