Climate Prediction with Tree Structure Based on Random Forest

Abstract

Climate prediction refers to the use of scientific methods and techniques to predict and analyze changes in the natural environment at a certain point in the future. This paper aims to improve climate prediction by analyzing its parameters and impacts using Random Forest (RF) techniques. Specifically, firstly, data frames (DF) are manipulated to explore data features, perform visualization, and perform preprocessing. Second, the paper introduces RF as a base model as a backbone network to accurately estimate climate change. RF is robust to outliers in the data using a tree-based approach. By integrating multiple decision trees and introducing stochasticity (e.g., randomly selecting features) during the training process, RFs are effective in reducing the risk of overfitting and improving the model's generalization ability. Third, mean absolute error loss (MAPE loss) is used to compare the errors that are evaluated on the data. The experimental outcomes demonstrate the efficacy of the proposed model in climate prediction, with the accuracy increased to 94%. Applying the proposed model in climate prediction in this paper provides valuable insights for climate prediction.