Forecasting onion armyworm using tree-based machine learning models

In the Philippines, the province of Nueva Ecija produces fifty-four percent of its annual onion production. However, the level of onion growth production was reduced; since the outbreak of 2016, armyworms destroyed thousands of hectares of farms resulting in a loss of billions of pesos, which lead to the decline of the onion harvest. In this study, we develop machine learning models to forecast an outbreak of armyworms to help evade or reduce the damage caused by an armyworm outbreak. Climatic data; particularly Maximum temperature, Minimum Temperature, Ultraviolet Index, Humidity, Cloudiness, Wind Speed, Sun Hours, Rainfall, and Pressure from the Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) and armyworm outbreak occurrences data from the Provincial Agriculture Office (PAO) of Nueva Ecija was used as the dataset for this study Using Tree-based machine learning models Decision Tree and Random Forest. Binary classifiers were developed and evaluated to forecast the occurrence or non-occurrence of the armyworm outbreak and the use of feature importance to distinguish the most critical climatic features that significantly contribute to forecasting an armyworm outbreak in the province of Nueva Ecija. These tree-based models produced satisfactory results, with the Random Forest model exhibiting a better forecasting capability than the Decision Tree model.