PEST CLASSIFICATION AND PREDICTION: ANALYZING THE IMPACT OF WEATHER TO PEST OCCURRENCE THROUGH MACHINE LEARNING

In the Philippines, pest infestation is the most common problem that affects rice production and the profits of our farmers. Anticipation may mitigate possible damages that a pest can bring. This study aims to classify and predict pests with the use of machine learning techniques. This will also attempt to identify the contributing variables that have an impact in terms of its accuracy. Prediction is based on the values of different weather parameters during infestation. Datasets containing weather variables were taken from an Automated Weather Station located in the town where the recorded infestation happened. Stem borer and brown plant hopper are the most common pests that infected the area and were selected for the study. Machine learning techniques such as Random Forest, Naive Bayes, XGboost, and CART were utilized and compared. This study also attempts to identify the factors affecting the growth of the two mentioned pests. It will determine the probability of these pests developing on a specific value of a particular weather parameter. It will provide a scientific basis for our farmers in terms of anticipation of a possible pest infestation. It showed that Random Forest and XGBoost algorithms performed excellently in terms of classification and prediction. For most machine learning techniques used, solar radiation and dew point were found to have a higher influence on affecting the accuracy of the model in performing prediction and classification. For future studies, results may be used and served as a basis for developing an Integrated Pest Management System and Decision Support System.