Using nondeterministic learners to alert on coffee rust disease

Abstract

Motivated by an agriculture case study, we discuss how to learn functions able to predict whether the value of a continuous target variable will be greater than a given threshold. In the application studied, the aim was to alert on high incidences of coffee rust, the main coffee crop disease in the world. The objective is to use chemical prevention of the disease only when necessary in order to obtain healthier quality products and reductions in costs and environmental impact. In this context, the costs of misclassifications are not symmetrical: false negative predictions may lead to the loss of coffee crops. The baseline approach for this problem is to learn a regressor from the variables that records the factors affecting the appearance and growth of the disease. However, the number of errors is too high to obtain a reliable alarm system. The approaches explored here try to learn hypotheses whose predictions are allowed to return intervals rather than single points. Thus, * Corresponding author Email address: antonio@aic.uniovi.es (Antonio Bahamonde) May 18, 2010 in addition to alarms and non-alarms, these predictors identify situations with uncertain classification, which we call warnings. We present 3 different implementations: one based on regression, and 2 more based on classifiers. Preprint submitted to Expert Systems with ApplicationsThese methods are compared using a framework where the costs of false negatives are higher than that of false positives, and both are higher than the cost of warning predictions.

An Empirical Multi-classifier for Coffee Rust Detection in Colombian Crops

“…Meaning that plants are vulnerable to rust attack, depending on environmental conditions and crop agronomy [2]. Since coffee rust has led to considerable losses in the industry worldwide, recent Brazilian supervised learning researchers have focused on detection of the incidence of the disease using simple classifiers as decision trees, support vector machines and bayesian networks [3,4,5,6,7,8]. They made use of numerical values of the infection rates which were mapped into two categories (or classes).…”

Toward detecting crop diseases and pest by supervised learning

“…Similarly, in [39] the research presented above is extended, using nondeterministic classifiers (carrying out modifications on the SVM regression algorithm) in order to predict if the percentage of coffee leaves infected by rust is above a threshold defined by an expert on the subject. The research in this way contributes to reducing the use of chemical fungicides, investment costs, environmental impact, and thus increases the quality of coffee production.…”

Interval prediction for graded multi-label classification

“…The central idea is that nondeterministic classifiers return more than one class when there are reasonable doubts about the right prediction, instead of risking a single prediction. These classifiers were introduced in Alonso et al (2008);del Coz et al (2009);Luaces et al (2011) for multi-class and for ordinal classification tasks, although these approaches are not devised to deal with multi-label data.…”