Valuation of Intangible Fixed Assets Using Generative Artificial Intelligence and Machine Learning

抄録

Despite the increase in the economic value of intangible assets in recent years, a problem remains regarding the inadequate reflection of these assets in corporate balance sheets. Moreover, although machine-learning models to evaluate intangible fixed assets exist, their versatility remains inadequate for dealing with complex models. This study aims to develop a new methodology for identifying companies and industries likely to have unrecorded intangible fixed assets on their balance sheets. Synthetic data were generated using the financial statement data of companies listed on the Prime Market of the Tokyo Stock Exchange, employing generative artificial intelligence (AI) techniques, following which a regression model was constructed based on machine learning. This study uses synthetic data to imitate the latent features and patterns of the real data, thereby providing new insights into the valuation of intangible assets. Generative AI techniques were employed to generate a large synthetic dataset that was applied to train a machine-learning model, and enhance its predictive accuracy and generalization ability beyond the limitations of the real data. Explainable AI techniques were also applied to increase model transparency and interpretability, allowing experts and general stakeholders to easily understand the predictive results of the model. The results suggest that intangible fixed assets are more likely to have higher values than their recorded values in certain industries. This study showcases the applicability of AI technology to financial analysis, specifically to the accurate pinpointing of the presence of unrecorded intangible fixed assets. Such an application can potentially provide additional information for investors and creditors when making investment decisions, and may contribute to decision-making effectiveness and a true understanding of corporate value. Future researchers are suggested to improve the model’s intangible asset valuation accuracy through dataset expansion and model refinement.