Neural network techniques for monotonic nonlinear models

Given a set of statistical data, one may use traditional regression models, such as linear regression and polynomial regression model, to determine a regression curve based on an assumption of the form of the regression equation. However, in cases where the relationship between the predictors (independent variables) and the response (dependent variable) is irregular, the regression results are often questionable. In this paper, a concept of monotonic nonlinear regression model is introduced. The neural network models used for monotonic nonlinear regression model are briefly discussed. An algorithm for monotonic nonlinear regression model is proposed. An example with real data in regression analysis using the neural network regression model is demonstrated. A summary of computer simulation experiments is presented.