Evolutionary Algorithm‐Based Radial Basis Function Neural Network Training for Industrial Personal Computer Sales Forecasting

Abstract

Forecasting is one of the crucial factors in applications because it ensures the effective allocation of capacity and proper amount of inventory. Because Box–Jenkins models using linear forecasting have their constraint to predict complexity in the real world, other nonlinear approaches are developed to conquer the challenge of nonlinear forecasting. With the same goal, we are proposing a hybrid of genetic algorithm and artificial immune system (HGAI) algorithm with radial basis function neural network learning for function approximation and further applying it to conduct an industrial personal computer sales forecasting exercise. In addition, five well‐known benchmark problems were used to evaluate the results in the experiment, and the newly proposed HGAI algorithm has returned better results than the Box–Jenkins models and other algorithms.

An approximate optimization strategy using refined hybrid metamodel

“…The kriging is a kind of interpolation that comes from geostatistics, and design and computer experiments (DACE) is well known for the application of computer simulation . The NN is a type of machine learning modeled on the human brain, and RBF uses the radial basis function as an activation function . In addition, there is a regression technique called multivariate adaptive regression spline in the field of statistics, which is a nonparametric regression technique that constructs multiple linear regression models across the range of independent variables .…”

“…[2][3][4] The NN is a type of machine learning modeled on the human brain, and RBF uses the radial basis function as an activation function. 5,6 In addition, there is a regression technique called multivariate adaptive regression spline in the field of statistics, which is a nonparametric regression technique that constructs multiple linear regression models across the range of independent variables. 7,8 When the variables have uncertainties, we can predict the function relation between independent and dependent variables through fuzzy regression analysis.…”

A hybrid FAM‐CART model for online data classification

“…Over the years, a number of data mining techniques have been proposed to solve real‐world classification problems . In this aspect, neural networks, such as the radial basis function (RBF) networks and multilayer perceptron (MLP), are useful models for undertaking data classification tasks . However, 1 key problem is that they are offline learning models and need retraining when new data samples are presented.…”

Development of a coupled wavelet transform and evolutionary Levenberg‐Marquardt neural networks for hydrological process modeling

“…There are several studies on the applicability of evolutionary neural networks (ENNs) in hydrology that have been presented in recent years . Among these works, GA more than others has shown great progress in training ANN models …”

“…[5][6][7] Among these works, GA more than others has shown great progress in training ANN models. 3,[8][9][10] Due to the local convergence of BP, the solutions are highly dependent upon the initial random draw of weights. If these initial weights are imposed on a local grade, which is probable, BP algorithm will likely be trapped in a local solution that may or may not be the global solution.…”