Associação de bioinsumos e nitrogênio no trigo: desempenho agronômico avaliado por métodos estatísticos e aprendizado de máquina

Abstract

The combination of biological inputs and nitrogen is a management strategy that still requires further study to assess its technical viability in improving wheat yield. To obtain more reliable results from the dataset, integrating conventional statistical analyses with machine learning approaches can be a valuable strategy. The aim of this study was to evaluate the relationship between foliar-applied biological inputs and topdressed nitrogen fertilization on the morphophysiological traits and performance of wheat, using both classical statistical tools and machine learning techniques. The experiment was carried out under field conditions at the UTFPR Experimental Station 3 Dois Vizinhos Campus, during the 2022 winter season, using the TBIO Toruk® cultivar. The treatments consisted of applying biological inputs at the beginning of tillering, as follows: T1 3 control; T2 3 Azospirillum brasilense (strain AbV5); T3 3 Bacillus velezensis (UFV3918); and T4 3 Azospirillum brasilense (AbV5) + Bacillus velezensis (UFV3918). Half of the plots in each bioinput treatment received 75 kg ha⁻¹ of topdressed nitrogen. At flowering, leaf area index (LAI) and photosynthetically active radiation (PAR) were measured. At harvest, the harvest index, thousand-grain weight (TGW), test weight (TW), and grain yield were evaluated. The experiment followed a randomized block design. Analyses were performed using Genes software and Python. The combination of biological inputs with topdressed nitrogen improved the morphophysiological and productive performance of wheat, with the Azospirillum + Bacillus treatment showing the best results. The integration of traditional statistical methods with machine learning helped identify patterns and demonstrated the potential of this approach to support decision-making in more efficient management strategies.