Variance and model optimization in sensor networks

This article describes suggested improvement of model's reliability using its statistical characteristic variance. At first a new model is constructed from available type and group of basis functions (radial and polynomial), its parameters are optimized to minimize the error and variance is calculated. The minimization of variance results in a location of an additional data point which will bring the best improvement of reliability. This approach should ensure that the model found represents the real situation better than finding a model solely with the least-squares criteria.