An efficient method of evaluating the distance between two uncertain objects

When data mining techniques are applied to uncertain data, their uncertainty has to be considered to obtain high quality results. Usually, an uncertain object is described by a probability density function, a probability density function is approximated by a large amount of sample points, and the distance between two uncertain objects is expressed by the expected distance. Computing the expected distance is costly because it involves double integral using a large amount of sample points for two uncertain objects´ probability density functions. This is critical for some uncertain data mining techniques. In this paper, a simple and efficient formula of evaluating the distance between two uncertain objects is presented. We also give the application of the formula in nearest-neighbor classifying. Experiments with datasets based on UCI datasets and the plant dataset of “Three Parallel Rivers of Yunnan Protected Area” verify the formula is effective and efficient.