D3398: Analyzing and estimating insect population dynamics using machine learning models with effective environmental data

Population dynamics of insect pest is closely related with their effective environment. In previous studies, abiotic factors (e.g. temperature, moisture, etc.) data have been used to estimate population dynamics. However, population dynamics is affected not only by abiotic factors but also by biotic factors (e.g. host, predator, etc.). Therefore, a method is needed to estimate population density with environmental factors those directly or indirectly affect density fluctuations. Machine learning makes a model by learning patterns by the computer itself based on the input data. Recently, as the amount and range of data are increasing, it is in the spotlight because accurate and fast information processing is needed. Machine learning estimates insect pest density by finding a relationship between environmental factors and insect pest density.

This study was conducted from March 2022 to May 2022 in pepper greenhouse in Jinju city, Gyengnam-province. Adults of Frankliniella occidentalis collected using visual survey. Environmental factors affecting density fluctuation of insect pests are selected temperature and humidity of the atmosphere, potential, and rate of photosynthesis and NDVI (Normalized Difference Vegetation Index) of the host, temperature, moisture, and EC of soil. These were measured weekly. Analyzed using KNN (K-Nearest Neighber) and Bagging (Bootstrap aggregating) during machine learning.