MedProcure: An AI-Enabled Hospital Procurement Management System with Multi-Level Approval Workflow

Abstract

Hospital procurement involves purchasing medicines, equipment and consumables while keeping stock available, spending controlled and every decision accountable. In many hospitals this process still relies on manual requests, informal approvals and periodic stock checks. That leads to delayed purchases, understocking of critical items, overstocking that wastes budget, and poor traceability. To address these problems, MedProcure, a web-based Hospital Procurement Management System, has been developed. It manages the full procurement lifecycle: purchase request, configurable multi-level approval, purchase order, supplier response, goods receipt and inventory update. The system has a secure role-based access control mechanism, master data management for items, suppliers, manufacturers, categories, units and departments, and an approval matrix that routes each request by department, category and amount. The platform is built with Angular for the user interface and ASP.NET Core Web API for backend processing. Data is stored in a MySQL database accessed through Entity Framework Core, and JWT authentication protects the system. To make procurement more intelligent, the system is being extended with machine learning modules. An inventory understock and overstock analysis module uses XGBoost to classify stock status and recommend procurement actions. A seasonal demand prediction module uses the Temporal Fusion Transformer, a time-series model that handles seasonality and shows which variables influenced a forecast. A smart purchase request review module uses LightGBM to compare a requested quantity against current stock and recent consumption and then recommend an appropriate quantity. The AI modules only recommend, and the human approver keeps the final decision. Together, these features aim to reduce stockouts and excess inventory, speed up approvals and support better decisions in hospital supply management.

Introduction

The text presents MedProcure, a web-based Hospital Procurement Management System integrated with Machine Learning (ML) to improve the efficiency, transparency, and accuracy of hospital purchasing and inventory management.

Hospitals require a continuous supply of medicines, medical equipment, and consumables. Traditional procurement processes often rely on manual requests, informal approvals, and periodic stock checks, leading to delayed purchases, shortages of essential items, excess inventory, and poor accountability. MedProcure addresses these problems by digitizing the entire procurement process and integrating AI-based decision support.

1. Key Features of MedProcure

The system provides the following functionalities:

• Master Data Management: Maintains consistent records of items, suppliers, manufacturers, categories, units, and departments.

• Purchase Request Management: Allows departments to submit purchase requests with required quantities and relevant details.

• Multi-Level Approval Workflow: Routes requests to the appropriate approvers based on department, item category, and purchase amount.

• Purchase Order Management: Converts approved requests into supplier purchase orders and records supplier responses.

• Goods Receipt and Inventory Updates: Records received goods and automatically updates inventory levels.

• Role-Based Access and Security: Uses JWT authentication to protect the system and control access according to user roles.

• Audit Trail: Records procurement activities and approval decisions to ensure transparency and accountability.

2. Technology Stack

MedProcure uses a layered architecture consisting of:

• Angular: Frontend user interface.

• ASP.NET Core Web API: Backend processing and REST API services.

• Entity Framework Core (EF Core): Database access and management.

• MySQL: Storage for master data, procurement transactions, approval rules, and stock records.

• JWT Authentication: Secure access to the system.

The backend follows the Controller–Service–Repository design pattern to improve maintainability and support future extensions.

3. Machine Learning Modules

The proposed system integrates three ML modules to support purchasing decisions.

A. Stock Analysis Using XGBoost

XGBoost analyses inventory information and classifies items as understocked or overstocked. Based on the results, it recommends whether to order additional stock or postpone purchasing.

B. Seasonal Demand Prediction Using Temporal Fusion Transformer (TFT)

TFT forecasts demand over multiple future periods while considering seasonal patterns and other relevant variables. It also provides insights into the factors influencing its predictions, helping approvers understand expected demand.

C. Smart Purchase Request Review Using LightGBM

LightGBM evaluates purchase requests by comparing requested quantities with current stock levels and recent consumption. It recommends appropriate quantities to help prevent unnecessary purchases and excessive inventory. Separate rule-based checks identify duplicate purchase requests.

4. Methodology and Workflow

The system follows the complete hospital procurement lifecycle:

1. Maintain master data for items, suppliers, and departments.

2. Create and submit purchase requests.

3. Route requests through the configured approval hierarchy.

4. Review requests using ML-based recommendations.

5. Convert approved requests into purchase orders.

6. Record supplier responses and goods received.

7. Update inventory and use the resulting data for future stock analysis and demand forecasting.

This integrated approach combines workflow automation with data-driven recommendations rather than treating inventory forecasting as a separate activity.

5. Research Contribution and Significance

Previous research has primarily focused on demand forecasting and inventory replenishment for specific categories of hospital items. MedProcure aims to bridge the gap by integrating procurement requests, approvals, supplier ordering, goods receipt, inventory tracking, and machine learning into one auditable platform.

Conclusion

MedProcure presents a web-based Hospital Procurement Management System that manages the full procurement lifecycle — purchase request, configurable multi-level approval, purchase order, supplier response, goods receipt, and inventory update — on a single platform. Its Angular frontend, ASP.NET Core Web API, MySQL database with EF Core, JWT authentication, and role-based access control provide a secure and maintainable foundation, while the approval matrix routes each request by department, category, and amount. The planned XGBoost, Temporal Fusion Transformer, and LightGBM modules extend the system with stock status analysis, seasonal demand prediction, and purchase request review, with the human approver retaining the final decision. Future work will complete the integration of these modules and evaluate them on hospital data to measure their effect on stockouts, excess inventory, and approval time.

References

[1] M. I. Fernandez, P. Chanfreut, I. Jurado, and J. M. Maestre, “A Data-Based Model Predictive Decision Support System for Inventory Management in Hospitals,” IEEE J. Biomed. Health Inform., vol. 25, no. 6, pp. 2227–2236, 2021, doi: 10.1109/JBHI.2020.3039692. [2] S. Petchrompo et al., “A Data-Driven Framework for Vaccine Demand Forecasting and Inventory Simulation in a Hospital Travel Clinic,” IEEE Access, vol. 14, 2026, doi: 10.1109/ACCESS.2026.3685781. [3] T. Chen and C. Guestrin, “XGBoost: A Scalable Tree Boosting System,” in Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discovery Data Mining, San Francisco, CA, USA, 2016, pp. 785–794, doi: 10.1145/2939672.2939785. [4] G. Ke et al., “LightGBM: A Highly Efficient Gradient Boosting Decision Tree,” in Proc. Adv. Neural Inf. Process. Syst. (NIPS), vol. 30, Long Beach, CA, USA, 2017, pp. 3146–3154. [5] B. Lim, S. Ö. Ar?k, N. Loeff, and T. Pfister, “Temporal Fusion Transformers for Interpretable Multi-Horizon Time Series Forecasting,” Int. J. Forecasting, vol. 37, no. 4, pp. 1748–1764, 2021, doi: 10.1016/j.ijforecast.2021.03.012. [6] M. Jones, J. Bradley, and N. Sakimura, “JSON Web Token (JWT),” RFC 7519, Internet Engineering Task Force, May 2015.

Copyright

Copyright © 2026 Sabarinathan R S, Udhaya Kumar M G, Sathya Narayanan K S, Sangamithra S. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.