Cascade Backward Propagation Neural Network and Multiple Regression in the Case of Heteroscedasticity

Abstract

Aims/ Objectives: To develop a new model called cascade backward propagation neural network performance over a filtered data by clustering algorithm based on robust measure (CFBNFDCARM). The performance of the clustering based neural network approach will be compare with the performances of regression analysis when the data deviate from the assumption of homoscedastic regression. Methodology: The new developed model was tested using the Airfoil, Aboline and Airline passenger data sets obtained from the UCI machine learning repository in order to compare the performances of regression analysis and a clustering based neural network approach when the data deviate from the assumption of homoscedastic regression. An algorithm based on robust estimates of location and dispersion matrix that helps in preserving the error assumption of the linear regression was introduced in the clustering technique. Results: The comparison indicated that the results emerging from our developed model gives a better performance when compared with the weighted least square regression as well as the standalone cascade backward propagation neural network for all the data sets considered. Conclusion: Analysis of the result showed that, the mean square error (MSE) and the root mean squared error (RMSE) in all the cases considered in this study decreases in a definite manner. From the obtained result, it can be seen that, our proposed model (CBPNFDCARM) performed better and can be a better alternative in dealing with heteroscedasticity in data set than both the weighted least square (WLS) and the standalone cascade backward propagation neural network (CBPN).