Preliminary Machine Learning Model for Citrus Greening Disease (HLB) Prediction in Colombia

Abstract :

Citrus greening disease (Huanglongbing-HLB) is considered the most destructive of citrus worldwide. Of the three species of Candidatus liberibacter associated with HLB, two have been recently reported in Latin America. This paper, adults from Diaphorina citri Kuwayama (Hemiptera: Liviidae) and material from suspicious citrus plants were collected from 21 departments of Colombia. The molecular detection, from 16S rDNA marker, by means of the real-time PCR test for the detection of the bacterium causing HLB, was validated for the monitoring of the disease in Colombia. In this study, a machine learning models is used to identify of the citrus greening disease (HLB). Finally, a comparison of the performances obtained by the algorithms Random Forest and K-Nearest Neighbors (KNN) is presented. The algorithm with the highest AUC value was Random Forest with 100%.