Two original procedures for clustering spatially uncertain data are proposed in this dissertation. The first procedure enables the clustering of spatial data based on bisector division of space, using comparison and cluster pruning. It significantly reduces the number of the expected distances calculations and speeds up the process of clusters pruning in comparison to existing procedures. Second procedure divides the data set area using spatial relations among objects to increase the possibility of parallel processing. The procedure enables parallel execution of the clustering process. The procedure does not require additional investments in equipment, because of use a computer with multiple cores. Presented procedures are used for creating a model for prediction of behaviour service-oriented system, using clustering of existing data about objects of service-oriented system. This model can predict the requirements for service-oriented system and prepare according to the requirements. Costs are reduced and increased the number of tasks that can be done by service-oriented system.