Uncertainty-based Spatial Data Mining

Although uncertainties exist in spatial data mining, they have not been paid much attention to. Uncertainty may have an influence on the confidential level, supportable level, and interesting level of spatial data mining. This paper proposes uncertainty-based spatial data mining. First, the concept is given in the integrated contexts of both uncertainty and spatial data mining. The inherent uncertainties that have their own characteristics play an important role in spatial data mining. Second, the external aspects and their internal sources of uncertainty-based spatial data mining are given. Besides the errors, spatial uncertainties further include positional uncertainty, attribute uncertainty, topological uncertainty, inaccuracy, imprecision/inexactitude, inconsistency, incompleteness, repetition, vagueness, noisy, omittance, misinterpretation, misclassification, abnomalities and knowledge uncertainty. Given a mathematical interpretation, the internal sources may be randomness, fuzziness, blunders, chaos, etc. To control and reduce uncertainty in an acceptable degree, one is data acquisition that highlights the information acquired from the process of data collection and data amalgamation, the other is data cognition that emphasizes the knowledge discovered from data extraction process and information generalization. Third, the usable techniques and methods that may possibly cope with the uncertainties in spatial data mining are briefly overviewed. For example, GIS data models, analysis of error propagation, probability theory and mathematical statistics, extended sets. The cloud model integrates the randomness and fuzziness by using the formalization-computerized language, and it is more appropriate when there exist more than one uncertainty at the same time, e.g., randomness and fuzziness. Finally, a case study is given on Baota landslide.