Committee of SGTM Neural-Like Structures with RBF kernel for Insurance Cost Prediction Task

A new method for constructing a committee based on the use of a set of SGTM Neural-Like Structures with RBF kernel for solving regression tasks was developed. The use of the RBF kernel for the hybridization of SGTM Neural-Like Structure allows increasing the accuracy of the method, and building a committee on their basis provides results that are more efficient. Modelling of the method occurred on the real data for the task of the insurance business. The numerical values of the accuracy of the method and the speed of training procedures in comparison with the existing ones are given. The highest accuracy of our prediction method in comparison with the existing ones is established. The developed committee can be used for solving various regression and classification tasks in the field of the insurance business. It is focused on solving tasks of the large dimension. Among the possibilities to increase its efficiency is the possibility of the hardware implementation of this committee with the parallelism of processes, when for each data cluster used its processor.