Mining Spatio-Temporal Datasets: Relevance, Challenges and Current Research Directions
M-Tahar Kechadi, Michela Bertolotto, Filomena Ferrucci, Sergio Di · 2009
This chapter discusses available techniques and current research trends in the field of spatiotemporal data mining. An overview of the proposed approaches to deal with the spatial and the temporal aspects of data has been presented. Approaches that aim at taking into account both aspects were also surveyed. Many challenges are still to be addressed. In particular, in the cartography and GIS community background/domain knowledge plays a significant role in the analysis of data. Therefore one of the biggest challenges is to integrate background geographic knowledge within the mining process. This is still an unexplored issue. Currently, huge volumes of data are collected daily are often heterogeneous, geographically distributed and owned by different organisations. For example, an application by its nature is distributed such as an environmental application for which the data is collected in different locations and times using different instruments, and therefore these separate datasets may have different formats and features. So, traditional centralised data management and mining techniques are not adequate anymore. Distributed and high performance computing knowledge discovery techniques constitute a better alternative as they are scalable and can deal efficiently with data heterogeneity. So distributed data mining has become necessary for large and multi-scenario datasets requiring resources, which are heterogeneous and distributed. This constitutes an additional complexity to spatio-temporal data mining. We will look at this problem in our ADMIRE framework (Le-Khac 2006). Visual techniques are essential for effective interpretation of mining results and as support to the mining process itself. We have discussed such techniques and presented our work in the area.