Evaluation the Efficienty of Forecasting Sports Events Using a Cascade of Artificial Neural Networks Based on FPGA

Abstract

Well-known implementations of software systems for predicting sports events based on artificial neural networks do not provide sufficiently accurate results in individual sports. The main difficulty in this case is the fact that various types and formats of data act as source information: numeric, symbolic, interval. The authors propose an approach based on cascading of several types of neural networks. A specialized training sample is prepared for each module in accordance with the type of model and the selected cascade structure. The results of information processing on each layer serve as data sets for modules of subsequent tiers. Initially, the system was implemented using the MATLAB package. The experiments carried out confirm the effectiveness of cascading modules. To significantly increase the performance of the system, for example, when used in betting shops, a cascade of neural networks is implemented on the basis of Altera Cyclone III FPGA using Quartus CAD.