Modeling for System Optimization with Small Dataset Using Neural Network

Abstract

The shortage of data is one of the most important problems in system modeling and optimization in industrial applications. Typical modeling techniques are unable to properly model a system with a limited dataset. In this paper, a modeling method for optimization of these systems is proposed. The proposed method has two main steps. In the first step, the model is employed to generate data using neural network. This model determines the correspondence input of each output. In the second step, optimization of the generated model is performed using genetic algorithm. Inputs leading to the specified output can be estimated using the proposed system. Optimality of the system can be explained by an evaluation function. The proposed method was evaluated in two different experiments on a time series and a real data. Results of the experiments were analyzed using mean square error. The experimental results show the capability of the proposed method in system modeling and optimization.