Volume 23, Issue 2 (6-2026)
jor 2026, 23(2): 151-176 |
Back to browse issues page
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Naji Azimi Z, Bargchi F, Razavi H, Estiri M. Predicting the Number of Packages Arriving at the Post Office Using Time Series, Regression, and Neural Networks. jor 2026; 23 (2) :151-176
URL: http://jamlu.lahijan.iau.ir/article-1-2316-en.html
URL: http://jamlu.lahijan.iau.ir/article-1-2316-en.html
Department of Management, Faculty of Economics and Business Administration, Ferdowsi University of Mashhad, Mashhad, Iran , znajiazimi@um.ac.ir
Abstract: (106 Views)
Demand forecasting is a key component in management decision-making, because customer demand for goods and services is not constant over time and fluctuates under the influence of various factors. This issue is more important in government organizations; in this regard, the post office, as one of the government service organizations, is directly faced with high demand volumes and significant fluctuations, especially in the field of postal parcel distribution. Therefore, providing an accurate and reliable forecasting model for the number of incoming postal parcels plays an important role in improving human resource planning, operation scheduling, and cost control. Accordingly, the main objective of this research is to introduce and evaluate a suitable model for predicting the number of postal parcels arriving at the post office. The innovation of this research is in identifying and considering new event-driven and environmental variables such as the impact of pandemics including COVID-19 and its resulting restrictions, online sales festivals, exhibition events, and school reopening. In this research, the General Post Office of Khorasan Razavi Province was selected as the study case. Historical data related to the number of incoming postal packages for 18 consecutive months was extracted on a daily basis, and the forecasting process was carried out using three common forecasting approaches including time series analysis, regression analysis, and artificial neural network. The results of comparing the performance of the models show that the feedforward neural network model, when the weekend is removed from the data, has the lowest error rate compared to other models. After that, the ARIMA time series model and finally the regression model with time series components are ranked next.
Type of Study: Research |
Subject:
Special
Received: 2026/01/15 | Accepted: 2026/06/11 | Published: 2026/06/22
Received: 2026/01/15 | Accepted: 2026/06/11 | Published: 2026/06/22
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
