Neural Network Committees Optimized with Evolutionary Methods for Steel Temperature Control

摘要: This paper presents regression models based on an ensemble of neural networks trained different data that negotiate the final decision using optimization approach evolutionary approach. The model is designed for big and complex datasets. First, clustered in a hierarchical way then level cluster random choice training vectors several MLP are trained. At test phase, each network predicts output vector determined by weighing outputs particular networks. weights algorithm merge genetic programming searching error minimum some directions. system was used prediction steel temperature electric arc furnace order to shorten decrease costs production cycle.