Prototipo de sistemas inteligentes para la predicción de la demanda de productos en pymes

Abstract

Abstract: The objective of this study is to design and implement a prototype intelligent system for predicting product demand, focused on supporting logistics management and strategic decision-making in SMEs. The study arose from the problems associated with these organizations' limited capacity to anticipate demand, leading to deficiencies in operational planning, resource allocation, and inventory control. The methodology used is based on the CRISP-DM approach for data analysis and modeling and the SCRUM methodology for incremental system development. In the modeling phase, a comparison was made between various machine learning algorithms, selecting Random Forest for showing the best balance between accuracy and stability, as it obtained performance metrics of MAE=3.76, RMSE=4.91, and MAPE=21.24%, demonstrating adequate predictive capacity on historical sales data. The predictive model is integrated into a web system built with Python and Django, which performs the following tasks: loading and validating CSV files, processing predictions, displaying results, and exporting reports. System verification includes functional and non-functional tests, achieving 100% satisfactory functional flows and an average processing speed of 1.77 seconds in performance tests with 10 UC, action control, and sensitive data encryption. These results confirm that the developed system is a viable solution to support predictive demand management in SMEs.