<![CDATA[Digital startups are highly vulnerable to financial distress due to limited historical financial data and imbalanced datasets between healthy and distressed firms. These challenges reduce the accuracy of existing prediction models, hindering early risk detection for investors and policymakers. This study aims to optimize financial distress prediction in Indonesian digital startups by applying Generative Adversarial Networks (GANs) for financial data augmentation. GANs are used to generate synthetic financial data that replicate real-world distributions, particularly for the minority class, to balance the dataset. A quantitative experimental design was employed, comparing baseline and GAN-augmented models trained on financial ratios such as ROA, ROE, and DER. The results show that the GAN-augmented model achieved higher accuracy (92%), precision (91%), recall (88%), and F1-score (90%) compared to the baseline model. These findings confirm that GAN-based augmentation enhances model robustness and prediction reliability under limited data conditions. The study contributes to financial distress prediction literature by integrating deep learning with synthetic data generation, offering a practical tool for early detection of financial instability in digital startups and supporting data-driven risk management in Indonesia’s digital economy.]]>
Copyrights © 2025
