Stochastic Analysis of Single Server Queuing System with Recurrent Customer Arrival in a Banking System Using Machine Learning

Abstract

Uncertain waiting times in banking client lines may negatively impact customer satisfaction. Although the formula for waiting time provided by Little's Law in queue theory is general, it cannot be directly used to offer finite wait time estimate in practice. Examining potential predictors of waiting time is the primary goal of this research. The paper's implementation of the Artificial Neural Networks approach makes use of the Fast-Artificial Neural Network engine. Artificial Neural Networks were trained using Resilient Propagation. The input neurone was compared using both a time-series and a structural method. It was suggested that structural variables such as Queue Length and Head of Line length estimator variables be enhanced by averaging the length from prior intervals and the number of servers. We utilized an experimental approach to find the optimal configuration for the number of neurons in the input and hidden layers. This research found that compared to the time-series technique, the structural approach yielded more accurate estimates. The outcome is also improved when the combination of updated helper variables is used.