Multi-output Hybrid GA-NN with Adaptive Mechanism

Abstract

This research presents a hybrid Genetic Algorithm Neural Network (GA-NN) model to simulate the physical tests procedures of Medium Density Fiberboard (MDF). Data included in the model are related to MDF properties and its fiber characteristics. Multilayer Perceptron NN is a reliable supervised machine learning model. The model learns from seven inputs fed to the network to produce prediction of three targets. In order to avoid result from local optimum scenario, GA optimizes synaptic weights of the network towards reducing prediction error. The research used a fixed probability rates for crossover and mutation for hybrid GA-NN model. GA-NN model is further improved using adaptive mechanism to help identify the best probability rates for crossover and mutation. Fitness value refers to Sum of Squared Error, which is the accumulation of network error in the Output Layer. The population fitness distribution will guide best rates for each epoch. Performance comparisons are among three models; namely NN with Back Propagation (BP), hybrid GA-NN and hybrid GA-NN with adaptive mechanism. Results show the hybrid GA-NN model perform much better than NN model with back propagation optimizer. Adaptive mechanism in GA helps increase capability to converge at zero sooner than the ordinary GA.