Abstract:
The evaluation process of a mobile and web-based inventory management system
targets small and medium-sized enterprises that operate in Sri Lanka. Automatic stock
management along with supplier warning systems and stock prediction analytics
backed by machine learning formed the core features of the system design. The
developed system uses a multi-tier structure which combines Flutter-based cross-
platform frontend with .NET 8 and SQL Server backend and Python-based XGBoost
predictive module hosted by Flask. The system allows users to track inventory in real
time alongside automatic stock alerts in addition to supporting users through their
chatbot interface. The system's functionality and non-functional capabilities received
evaluation by both expert opinions and system testing procedures. The developed
system proves effective in improving stock control precision while cutting down supply
chain waste and simplifying procurement processes which establishes it as an ideal
solution for SMEs wanting to digitalize their inventory management systems.
The mobile and web-based inventory management system evaluation focuses on
assessing small to medium-sized enterprises that conduct business in Sri Lanka. The
system's main features incorporated automatic stock management while using supplier
warning systems and stock prediction analytics which operate through machine
learning technology. The developed system implements a multi-tier architecture that
unites Flutter-based cross-platform frontend with .NET 8 and SQL Server backend and
Python-based XGBoost predictive module hosted by Flask. Through its real-time
inventory tracking system the program provides automatic stock alerts to users who can
also find support from its chatbot interface. Expert opinions together with system
testing procedures evaluated the functionality alongside non-functional aspects of the
system. The developed system delivers precise stock control with waste reduction in
supply chains and operational ease so it represents an optimum solution for SMEs
creating digital inventory solutions. The three main technical components of the Inventory Management System deployment included a Flutter-based frontend application for mobile and web accessibility and a .NET 8 SQL Server backend system for secure data management through REST APIs. Additionally, the system incorporated a Python-based XGBoost predictive module for future stock analysis. The entire software relies on Entity
Framework Core, MediatR and AutoMapper frameworks to improve system organization while promoting concerns separation. The Flask API exposed the predictive model to the backend after model training occurred with real historical stock data processed through Pandas and NumPy. The application received new functionalities that included PDF invoice generation and JWT authentication and role- based access control while SMS notifications provided additional usability features. Standard data science metrics measured the effectiveness of the marketplace stock prediction model. The choice of XGBoost algorithm emerged as the optimal solution since it produced the best accuracy and efficient processing capabilities. The predictions were evaluated using retail inventory data which came from a dataset spanning multiple months. The implementation of the ML module within the operational system resulted in better demand prediction accuracy enabling staff to avoid unnecessary stockout or surplus situations according to expert evaluations and system tests.