Green finance growth prediction model based on time-series conditional generative adversarial networks

Area of Science:

• Environmental Finance
• Computational Finance
• Time Series Analysis

Background:

• Climate change mitigation requires substantial investment in green sectors.
• Accurate forecasting of green finance growth is crucial to encourage investment.
• Existing prediction models may lack accuracy due to data limitations.

Purpose of the Study:

• To propose and validate a novel methodology for predicting green finance growth.
• To enhance prediction accuracy through data augmentation techniques.
• To encourage increased investment in green sectors globally.

Main Methods:

• Utilized time-series Conditional Generative Adversarial Networks (CT-GANs) for data augmentation.
• Employed Nonlinear Autoregressive Neural Networks (NARNNs) for green finance growth prediction.
• Applied the methodology to datasets from forty countries across five continents, validating data non-stationarity with the Augmented Dickey-Fuller (ADF) test.

Main Results:

• NARNNs trained with CT-GAN augmented data achieved superior performance (R-squared: 98.8% Europe, 96.6% Asia, 99% other).
• CT-GAN augmentation significantly improved R-squared and RMSE compared to baseline NARNN models.
• The proposed model outperformed the Nonlinear Autoregressive Exogenous Neural Network (NARX-NN) across all regions.

Conclusions:

• The proposed methodology combining CT-GANs and NARNNs effectively predicts green finance growth.
• Data augmentation with CT-GANs is a valuable technique for improving time-series forecasting in finance.
• The findings support increased investment in green finance by providing more reliable growth predictions.