Neuro-fuzzy estimation in spatial statistics

Abstract

Although spatial statistics was developed based on probability and classic statistics, the data usually handled by them are frequently very approximate and linguistic and certainly are not suited for the probability concept. Furthermore, the traditional spatial statistics is developed principally for mining situations. When the approach is applied to problems under other situations such as air and water pollution, certain basic assumptions need to be modified. In an earlier paper, fuzzy spatial statistics was proposed. In this paper, neural learning combined with fuzzy representation is suggested for handling the variogram, which is essentially a covariance correlation, and the kriging, which is an unbiased method to estimate the missing data. Based on the fuzzy adaptive network, various computational methods are proposed to solve the resulting spatially distributed problem. © 2000 Academic Press.