Investigating Sensitivity of Nonlinear Classifiers by Reducing Mean Square Error (MSE)

Abstract

In this paper a new method is presented for handling the problems of Artificial Neural Networks (ANNs) in a Self-Organizing Map (SOM), the problems are the presence of noise and missing in datasets. Based on a robust and speedup of the investigation of a nonlinear classifications in SOM and Adaptive Linear Neuron ANNs (Adaline) algorithms that dealing with missing or noise data a new model proposed. the main objective of the proposed model is to contribute the advantages of all methods (SOM with Adaline) in rising the quality of SOM to handle the noise and missing value to improve its sensitivity of our model using combing the advantages of all methods (SOM with Adaline) in anew algorithm abbreviation (ADA-SOM). Simulation results show that the algorithm ADA-SOM achieved better performance and higher sensitivity to accurate MSE (Mean Square Error) than other standard classifiers, our model gets accurate benchmark results.