Comparison of the variable selection methods on prediction of sugarcane quality

• Grant number: 25/07385-2
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: June 01, 2025
• End date: May 31, 2026
• Field of knowledge: Agronomical Sciences - Agricultural Engineering
• Principal Investigator: Tatiana Fernanda Canata
• Grantee: Felipe Pontes Monachesi
• Host Institution: Faculdade de Ciências Agrárias e Veterinárias (FCAV). Universidade Estadual Paulista (UNESP). Campus de Jaboticabal. Jaboticabal , SP, Brazil
• Associated research grant: 23/12986-0 - Mapping sugarcane quality using artificial intelligence tools and high-resolution imagery, AP.R

Abstract

The methods of variable selection have been used in several contexts to solve challenges related to identifying patterns in datasets. Considering that the data obtained in agriculture tend to generate a significant number of variables that are not linearly correlated, there is a need to apply alternative methodologies for adequate treatment of georeferenced data. The objective of the project is to compare computational methods for selecting variables related to the sugar content of sugarcane plants. The methodology is based on statistical and machine learning tools to analyze data related to the vegetative crop vigor and the climatic conditions of the study area. Such tools will help to evidence which variables are associated with the sugar content of the plants so that can later be applied in predictive models aiming at greater accuracy in the spatialized prediction of the qualitative attributes of the crop. A commercial sugarcane area in the state of São Paulo will be used to obtain georeferenced samples in the field during the crop's maturation period, followed by laboratorial analysis of the samples. The database will consist of multispectral satellite imagery, meteorological data, and ground-truth data. Free software will be used to organize and geoprocessing the dataset, as well as to prepare the results in the form of graphs and thematic maps. The project results will highlight the types of integrated data that best compose the development of predictive models regarding the quality of raw materials and contributing with technical solutions that complement the conventional methods of data analysis in agriculture. (AU)