Spatiotemporal interpolation through an extension of differential evolution algorithm for agricultural insurance claims

Assessment of actuarial risks in agricultural insurance is a specific area which evaluates risks in geographical framework. In order to manage weather related risks, the agricultural insurance portfolio is analyzed according to spatial and temporal characteristics of hazard regions. To reduce the basis risk associated with the location of meteorological stations, inverse distance weighting method with reduction approach is employed to interpolate meteorological values related to the location and the time of the reported agricultural claims. The results are evaluated for the data provided by Turkish Agricultural Insurance Pool, which is a unique agricultural claims data set covering more than 100 products for 5 different hazard types. Since height is a very significant factor that determines the values for the meteorological variables concerned in this study, altitude values are used to choose the best initial sample set. We extend stochastic differential evolution (DE) optimization algorithm to a multivariate setting for the spherical space to measure the closeness of estimated altitudes to actual altitudes. Moreover, we propose an optimal way of choosing the population size for the initialization step as another improvement in relation to DE algorithm.