Aplicação de modelos estocásticos para previsão de demanda de um produto agroquímico

Abstract

Demand forecasting in the current market context is an almost essential tool. The competitiveness of contemporaneity implies the imposition of increasingly robust processes, with less subjectivity for the maintenance and survival of organizations. In this space, the present work, through quantitative, explanatory, and experimental research, aims to evaluate the behavior of time series forecasting algorithms in forecasting demand for a product of the agrochemical industry. To this end, the related literature is presented in the context of the agro-industrial market to cite and introduce the methods and algorithms with the highest degree of usage. In addition, it discusses demand forecasting in the agrochemical sector, presenting and applying the ARIMA methodology to a set of sales data for an herbicide product provided by a multinational company with a large presence in Brazil. It is possible to notice that the model used had little representation in the expectation of forecasting future demand given the behavior of waste curves. However, the established scientific method and the structuring of the process are, notably, positive results from this work.