There is an emerging focus on real-time data stream analysis and visualisation on mobile devices. The recent growth of smart phones capabilities has attracted many fields to port their applications to the mobile context. Examples of such mobile data analysis applications are patient monitoring, traffic monitoring, an stock market visualisation systems. Those types of applications require real-time analysis and visualisation of continuous, and voluminous datasets. Users of those systems need to make well-informed decisions in real-time based on the analysis of constant monitoring
on different sensors or other data sources. Thus, mining data streams on mobile devices is becoming increasingly an important field that is termed as Mobile/Ubiquitous Data Mining (UDM).
UDM refers to the process of data analysis that is performed on portable and embedded devices in distributed environments where data is transmitted in form of a continuous stream. Users, who are in a need for data stream analysis and visualisation in time critical scenarios, are the main target of UDM visualisation since it allows data mining to be accessible anywhere at anytime. However, visualising the outcome of UDM analysis on mobile devices faces many challenges due to the resources constraints and the characteristics of display screens of mobile devices as well as the distinctive characteristics of data streams. While limited computational resources, small screen space, and limited power supply are some of the limitations of mobile devices, data
streams also have challenging characterises such as being dynamic, changing, and continuous.
Several resource-aware and adaptive strategies and algorithms have recently been developed for UDM. This trend has shaded the lights on UDM visualisation as it is essential to quickly and appropriately understand the data analysis outcome. Clutter has been identified as a key issue in UDM visualisation, and clutter-adaptiv techniques have also been developed. However, visualising the continuous analysis of real-time data streams also requires adaptive, interactive, and user-driven style of presentation to ensure accuracy and appropriate comprehension of the data visualisation for mobile
Therefore, in this thesis, we propose, develop, and evaluate interactive, adaptive, generic, and user-driven/controlled UDM visualisation strategies that extend the state-of-the-art clutter-aware UDM visualisation strategies. We have implemented and
experimentally evaluated the proposed framework to examine its effectiveness in terms of clutter reduction, visualisation enhancement, and improving resource management.
The system was evaluated using publically available stock market data as well as geolocations dataset that was obtained by geo-coding addresses from the web. Finally, we have provided demonstrations of the application, which has been made available on the web and on the Google Android App-Store for free download to show the operation of our system. The outcomes of this dissertation have resulted in a Demo paper in ICDM 2010, and a conference paper in ICTAI 2010.