Research On Uncertainty Of Spatial Data Mining

The spatial information resource via spatial data mining have materially value in exploiture and use, therefore, they are the significant parts for continuable development decision-making in terms of economy, resource, environment, population, and society. However, the existence of spatial data mining uncertainty has a negative influence on the mining effect. Hence, it's necessary to work over the spatial data mining process and result. The uncertainties of spatial data mining hail from multi- uncertainty of spatial data itself and mining process, the stochastic uncertainty and fuzzy uncertainty are central points among them. In this paper, the two kinds of uncertainty from mining process will be studied.Here the main works are:(1) The origin, types and spread of spatial data mining uncertainty from spatial data and spatial data mining process are discussed thoroughly.(2) The essential of stochastic uncertainty from spatial data mining process is analyzed. And its transfer methods and synthesis methods are concluded by means of error theory and calculation. A discriminance of stochastic uncertainty based on error checkout is put forward. This discriminance can differentiate whether the stochastic uncertainty is aroused by system error or stochastic error effectively via example validation, it can also analyze what effect each factor would have on the emulational result and the experiment result. Thus, this discriminance will provide a positive function in the aspects of emulational software improvement, spatial data measurement methods selection and parameter correction of transfer and synthesis.(3) The fuzzy uncertainty from spatial data mining process is dealt with fuzzy dualistic relation. First of all, fuzzy dualistic relation is established in each phase and phases mutually. On the basis of its conception and operation, the property of "max-min" transfer is proved. Moreover, an algorithm combined with "max-min" synthesis methods of fuzzy dualistic relation and the synthesis illation of fuzzy rules under multi-estates is proposed, it can calculate the confirmed fuzzy estate circumscription. Thus, the uncertainty problem of mathematics expression can be transformed as certainty issues to deal with. The result and analysis of example shows the algorithm proposed can find the confirmed fuzzy estate circumscription exactly, and it also has a practicality in the approximate disposal of fuzzy uncertainty from spatial data mining process.(4) In this paper, a spatial data mining algorithm based on fuzzy clustering and fuzzy model discrimination is also proposed. The function of this algorithm is for the sake of reducing the uncertainty from spatial data mining process and achieving better mining outcomes. With the comparison of two emulational results, it shows this algorithm can reduce the negative influence arises by uncertain factors from spatial data and spatial data mining process. Therefore, the uncertainty is played down and the spatial data mining result would have more exactness and application value.