Use this URL to cite or link to this record in EThOS:
Title: BIG-CITY : common semantic models for big data applications in future smart-connected cities
Author: Gaur, Aditya
ISNI:       0000 0004 7655 3680
Awarding Body: Ulster University
Current Institution: Ulster University
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
The objective of the thesis is to present a common semantic approach to facilitate heterogeneous information fusion for activity recognition and a new rule inference process using probabilistic, machine learning and semantic-based methods. Initially, a multi-level smart city architecture is proposed in order to exploit heterogeneous information collected from different areas of a smart environment. Also, using different information fusion approaches, a generalized smart city ontology model is proposed that combines other domain data in order to deal with information uncertainty and data heterogeneity aspects of the smart city environment. The semantic model shows an improvement in prediction of an activity when the surrounding sensorised information is included during the activity recognition process. Further, an automated approach known as Combination Operator Selection Algorithm (COSA) is presented that infers automatically the correct form of mathematical combination operator to be associated with sensorised objects for an activity. The automatic approach helps in learning new rules that will be used in defining the knowledge for an activity inference process in a smart city environment. Further, a scalable solution to the automated approach above is presented using an association rule mining process. The novelty of the work lies in the inference of different forms of mathematical combination operators to be associated with sensorised objects for an activity using hidden association rules. Also, for mining the different mathematical combination operators an association rule mining solution is proposed that is faster than the earlier ECLAT (Equivalence Class Transformation) [137], [156]) association rule mining approach. The derived model based on the approach above motivated the development of an online activity inference solution that contributes to an improved activity recognition process. The ongoing activity prediction module uses a few sensorised objects combination arrangements in order to predict an activity. The model is applicable in situations where the dataset is of restricted size, and various activities take place in a common location during the same time period, and for which data are collected using a common set of sensorised objects. Later, an n-COSA method is proposed that utilizes COSA for inferring a user's next state based on knowledge of the user's previous n states. The n-COSA uses both density-based clustering and Apriori-based association rule mining to predict the user's next state. The density-based clustering approach is used to find the interesting user locations or Points of Interests; the Apriori approach is used to generate COSA operators and the sequence of associated, disassociated and independent user states corresponding to a user's spatial trajectory from GPS data. The n-COSA aims to be more precise and accurate in predicting a user's mobility pattern than the existing Mobility Markov Chain (n-MMC) model [50]. The research presented in this thesis contributes to a common semantic model defining different forms of mathematical combination operators with different levels of combination details in an activity ontology tree. The inferred semantic model is utilized to learn new rules and has the potential to enhance assistive services in a smart city environment.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available