Desenvolvimento do módulo de uma aplicação de business intelligence para análise de produtividade em fazendas verticais

Abstract

Agriculture plays a crucial role in ensuring food security and meeting the growing needs of the global population. However, the challenges inherent to conventional agriculture, such as its dependence on climatic factors and the pressure exerted on finite natural resources, instigate the search for more efficient and sustainable solutions. Thus, vertical farms emerge as a promising alternative for expanding food production in a controlled and optimized manner. Additionally, the analysis of significant volumes of data coming from sensors in automated indoor gardens, together with the application of business intelligence techniques, has the potential to increase the performance of vertical farms in relation to conventional agriculture, due to their ability to process large volumes of data, and transform them into insights that can help understand important variables for the performance of plant cultivation. This study addresses the development of a system and its application to generate predictions based on a set of data, using regression models such as Random Forest, Ridge, Lasso and Linear Regression. The results indicate a superior performance of the Ridge and Linear Regression models, with a Mean Square Error (MSE) of 3.09 and a Determination Coefficient (R²) of 0.74, for the set of data from Farm Weather Data, while for the simulated data set, it proves the robustness of the developed module, with the variance of MSE found, however remaining above 60% of R 2 . This contributes significantly to a deeper and more informed understanding of the prospects and effectiveness of these innovations in contemporary agriculture. In this way, the study made it possible, through a personalized system, to carry out descriptive analyzes and regressions, which could assist in the prediction and understanding of variables that can influence plant productivity.