Predictive Modeling of Queue Lengths and Waiting Times in E-Commerce Platforms: A Statistical Approach

Abstract

The ever growing e-commerce sectors have posed new problems to do with how to control for the number of customers as well as the waiting times especially during peak sales periods affecting customer loyalty. The purpose of this current research work is to construct predictive models for forecasting queue dimensions that occur in e-business environments in order to improve operations functioning. This research applies multiple modeling techniques ARIMA, Linear Regression model, Neural Network model, employed the hybrid modeling technique to compare the performance of different models in variance in digital traffic patterns. Log files of our platform were analyzed using data from six months, data was preprocessed and model was assessed by Mean Absolute Error (MAE), Mean Squared Error (MSE) and Co-efficient of-determination (R2). The statistics show that the best results are obtained by the Neural Network model, whose R￾squared is 0.85 during heavy traffic as well. We found that ARIMA achieved steady results for off-peak periods and that Linear Regression could be a decent benchmark. This paper’s conclusions indicate that it is possible to apply and implement adaptive predictive models that can help e-commerce queue management to improve the given situation more effectively and optimize customers’ waiting time. This study provides a systematic method for real-time traffic prediction of e-commerce which should open up directions for further examination in mixture and real-time evaluation models