Extracting knowledge from mobile users data

The adoption of mobile devices in our daily lives has provided the opportunity to better understand the social interaction behaviour of mobile users. Such understanding will enable practical applications in the formulation of commercial strategy, counter terrorism technologies and formulation of government policies. This thesis attempts to solve the problem set within a mobile environment that contains mobile phone users carrying mobile devices. The key challenges of the problem are to identify the groups that meet regularly, the physical locations that they travel to over time, and the thematic nature of the physical locations that they travel to. As technology advances allow for the current time and position of mobile users to be tracked precisely using systems such as the GPS, a valuable opportunity exists for business and government to extract intelligence from data recorded resulting from activities performed by mobile users. The problem that the thesis attempts to solve is the lack of solution in the situation where a large amount of movement data is available and the owners of such data need architectures and algorithms that provide the ability to extract patterns of interest in order for them to decide and better manage and allocate their valuable resources. This thesis describes 5 techniques in understanding the behaviour of mobile users through knowledge extraction from movement data. The first contribution is meeting pattern presents a set of mobile users that tends to meet more regularly than other sets of mobile users. Meeting pattern also addressed an inherent limitation as proposed in group pattern which presents a set of mobile users that meets regularly but influenced by the size of movement data available. 11 The second contribution is location segmentation technique in which the geographical area is segregated by using polygons to describe each and every area. It contributes to knowledge extraction process by grouping individual coordinates into a single area thus summarising the substantial nature of the area such as the library into a single area. This has a significant improvement over the knowledge extraction process as each area are no longer described in exact (x, y) coordinates instead being grouped into a single area. The third contribution is movement pattern in which the knowledge of how individual and groups of mobile users travels and interacts can be readily described. This tells a clear picture of their social interaction behaviour by identifying the groups that tends to travels as a single group. This adds value towards the meeting pattern as it reinforce the evidence that the meeting pattern group are indeed travelling together frequently instead of meeting regularly. The fourth contribution is Sliding Window technique which is a novel technique to be applied on a Location User Database (LUD) in order to quickly draw the knowledge out without the need for the process to analyse the whole LUD. In a data mining environment where the size of dataset could be enormous, this technique provides a realistic way in analysing dataset that are selected for knowledge extraction. This technique extracts knowledge on a real time basis. The fifth contribution is location thematication where each physical location is given a hierarchical naming structure in order to describe each of them in as much detail as possible. The design of the location the matication technique aligns with the Sliding Window and movement pattern techniques in efficiently providing knowledge of how individual and groups of mobile users travels in the form of what types of location of interest that they visit and interacts. The five contributions above represent a sequence of algorithms and techniques which will enhance the current domain of knowledge extraction from mobile user movement iii data. These contributions describe knowledge in the form of elements comprising mobile user, groups of mobile users, location, series of location visited and the inherent type of location visited. It encompasses the techniques in extracting and describing all types of knowledge that can be analysed from mobile user movement data.