Optimization And Applied Researching For Genetic Algorithm Based ART Network

To our human society,the earthquake caused tremendous losses.Using the current science and technology to study various factors of earthquakes and to identify the internal laws,artificial intelligence is a hot area of research in achieving the purpose of earthquake prediction.Through the neural network to study the earthquake has a broad research base.Through inspection of the relevant papers found that the adaptive resonance theory(ART) has been developed to avoid the stability-plasticity dilemma in competitive networks learning.The stability-plasticity dilemma addresses how a leaning system can preserve its previously learned knowledge while keeping its ability to learn new patterns.As an efficient categorizing network,the supervised ART network has wide use in the present as well as in the future.The different effects of input attributes on category results in supervised ART network was normally ignored by traditional researches.However those effects are quite important during the predictive stage in the applications.In this paper,we present a novel supervised ART network named Attribute Weight based ART (A WART) network.In AW ART network,we have defined the attribute weights and assign them to some of the nodes in the network.Therefore the different attributes will give different effects to the category results according to the attribute weight on the nodes of network.Meanwhile the Genetic Algorithm was taken to optimize attribute weight to improve the prediction accuracy.The paper also presents a rule extraction algorithm to extract the available IF-THEN rules from the trained network according to the network architecture.Earthquake prediction is a world-wide problem.A variety of factors bred the earthquake.In particular,the causes of earthquake are complex and highly uncertain. However,the neural network has a good structure of non-linear processing power,so the content of our research will be applied to forecasting the earthquake magnitude. This paper proposes a new method of learning and apply it to the earthquake prediction,what's more,proposed rule extraction method enhance the credibility of the neural network.To some extent,these studies of earthquake prediction has a small step to study this field.