APPLYING MULTIPLE CASCADE GENERALIZATIONS TO PREDICTING OBESITY LEVELS

Abstract

Cascade generalization is used as the main structure for predicting obesity levels. Cascade generalization (CG) is a sequential combination of machines, where the result of the previous machine is applied to the current machine. Two sets of CG are generated, the first set receiving true data and the second set receiving false data, to predict true and false obesity levels, respectively. The results of both sets are combined to find the best prediction result. In this paper, two types of machines are used to generate CG, namely, a single neural network with multiple outputs and a binary neural network with multiple outputs. The accuracy results obtained using the proposed multiple CGs were found to be better than those obtained using individual machines combining the proposed technique.