Bootstrapping the Multilayer Feedforward Propagation System for Predicting the Arrival Guest in Malaysia

Abstract

Predicting arrival guest is an essential step in estimating the Malaysia economic impact, particularly in 
short and long-term COVID-19 crisis. Neural Network family of models has been widely used in economic 
and tourism. Most of the study used single layer of fed forward propagation system, it is because of less 
sampling variation in Neural Network. However, apparently the multilayer provides a better predicting 
result but its disadvantage is to deal with high sampling variation. The motivation of this study is to enhance 
the ability of multilayer Neural Network in predicting the arrival guest. In this study, a hybrid model based 
on Bootstrapping the Neural Network proposed to predict the arrival guest of Singapore in Malaysia. The 
weights of variables to the first hidden layer nodes are bootstrapped. The subsequent hidden layer takes 
the bootstrapped weights in order to obtain the Neural Network output.  The prediction obtained by 
hybrid model has been compared with conventional multilayer Neural Network in terms of small variation. 
The computational results shows that the hybrid model provides better performance in predicting the 
arrival guest.