Proposta e implantação de uma solução de previsão de demanda em uma indústria de implementos agrícolas

Abstract

This study aimed to propose and validate a demand forecasting solution for an agricultural implements industry. A model-based forecasting process was developed according to the specific characteristics of the company under study. In addition to the process, a computational solution was implemented in the Python programming language, featuring an interactive interface in Streamlit and integrating both quantitative and qualitative forecasting models. The models were evaluated using the MAE and WAPE metrics, and the results were discussed with the management team. Forecasts were generated for the months of August, September, and October 2025. Within the scope of this research, it was possible to compare the forecasting errors of two out of the three forecasted months, as the actual results for October were not yet available. For August and September, the average overall WAPE was 6.69%, and the MAE was 16 units. The practical application of the solution produced positive impacts on purchasing processes, production planning, and the definition of sales targets, promoting greater integration among departments and reducing subjectivity in decision-making. The study also demonstrated the applicability of a hybrid approach that combines quantitative methods and expert judgment, showing its potential for replication in other companies within the sector.