Establishing the Potential Rice Loss Prediction Model of Climate and Nature Disaster Factors Based on Machine Learning Theory

Abstract

The United Nations points out that extreme climate events are frequent and widespread in the 21st century and have become a global security issue. Artificial intelligence and machine learning have attracted much attention in environmental applications. This study aims at applying machine learning (ML) to rice disaster prediction, and uses SPSS to analyze environmental impact factors. After model training and evaluation, four models are provided, among which short-term prediction results show high accuracy on a single event, which are suitable for water damage, cold damage, and plant diseases and insect pests respectively. In terms of longterm prediction, using future meteorological prediction values to predict potential rice losses is better, especially within a specific time period. Ultimately, relevant units such as the Council of Agriculture or the AgriFood and Food Administration can choose a suitable model based on different purposes (short-term or long-term forecasting).