UMine: Study on Prevalent Co-locations Mining from Uncertain Data Sets

Abstract

We can collect a large amount of spatial data by utilizing sensor positioning technology and wearable devices. However, most of the acquired data are uncertain because of the gaps in data collection or to maintain subject privacy. Thus, we investigate co-location pattern mining problem in the context of uncertain data. The prevalent co-location pattern under uncertain environments has two different definitions. The first definition, referred as the expected prevalent co-location, employs the expected interest degree of co-location to measure whether this pattern is frequent. The second definition, referred as the probabilistic prevalent co-location, uses the probabilistic formulations to measure frequency. Here a novel system called UMine is proposed to compare this two different definitions with a user-friendly interface. The core of a system such as this is the mining algorithm, and UMine is integrated with the expected mining method, probabilistic mining method, and approximate mining method. In this paper, the system is introduced in detail, and the comparison between these two types of definitions is implemented. The experimental results show that the difference between these two definitions’ result sets changes as the threshold changes. By flexibly adjusting the parameters, users can observe interesting patterns in the data. In addition, the demonstration provides data generation and preprocessing function while showing its practicality for either real-world or synthetic data sets. The study can also provide support for the further uncertain Co-location patterns mining research.