Detection of Rust Emergence in Coffee Plantations using Data Mining: A Systematic Review

Abstract

Hemileia vastatrix is a fungus that causes coffee rust disease and, depending on the level of severity, reduces the photosynthetic capacity of the plant and of new shoots, leading to low coffee yields and even death; its symptoms are visible on the leaf. Systems based on computer algorithms have been developed to predict diseases and pests in coffee. The objective of the manuscript was to analyse the detection of rust occurrence in coffee plantations, through field determinations of climatological, agronomic and crop management variables using data mining algorithms. A systematic review of studies published from 2001 to 2021 was carried out in the Scopus, Ebsco Host and Scielo databases, considering as an inclusion criterion the works that used experimental design in data collection. The studies included in this review were 22, 64% of which came from the top two coffee-roducing countries in Latin America (Brazil and Colombia); the analysis of these studies revealed that the input variables were climatic, soil fertility properties, management and physical properties of the crops. In addition, they used supervised (decision tree, artificial neural networks, multiple linear regression, among others) and unsupervised (clustering) algorithms, with the support of experts in the study of the fungus and used statistics such as coefficient of determination, root mean square error, among others, to validate the proposals. Overall, this systematic review provides evidence of the effectiveness of data mining algorithms implemented to detect the occurrence of rust in coffee plantations.