Role of green finance instruments in shaping economic cycles

Abstract

This paper evaluates the predictive performance of various machine learning models in economic forecasting using cross-validation and bootstrap bagging techniques. Focusing on a key area in economic forecasting, this study compares these models using cross-validation and bootstrap bagging techniques. The study uses a detailed dataset of green bonds issuing organizations from 30 regions of China from 2014 to 2022. The results indicate varying levels of efficacy among the models, with the deep multi-layer perceptron (DMLP) model showing better performance in accuracy and generalizability. When equipped with cross-validation, the k-nearest neighbor (KNN) model performed best among the five models. However, the decision tree is observed to be the best model when the bootstrap bagging technique is applied to all the five models. These findings highlight the potential of machine learning models to enhance economic forecasting accuracy, providing valuable insights for managers and economists in selecting suitable predictive models. The research contributes to understanding predictive modeling in economics, offering insights into applying machine learning techniques for accurate and reliable economic forecasting.