抄録
In recent years, despite the increase in the share of intangible assets in company assets, most are not recorded on their balance sheets. Therefore, intangible asset valuation methods have been studied widely. This study proposes a method for valuing intangible assets using machine learning, based on which a method to extract companies with a high probability of having intangible assets not recorded on their balance sheets has been verified. First, a model to predict intangible assets using machine learning was built. To construct the machine learning model, the data of the financial statements of 28 stocks in the electric machinery industry, forming the Nikkei 225, was used. Data on intangible assets were prepared as the target variable, and data on items listed in financial statements as features were prepared to build the model. A machine learning model was constructed by a regression analysis using the holdout method, and the model was evaluated using the coefficient of determination. Then the valuation of intangible assets was estimated using this machine learning model. After creating a scatter plot in which the valuation of intangible assets estimated by the machine learning model was plotted on the horizontal axis, the valuation of intangible assets recorded on the balance sheet was plotted on the vertical axis. The results revealed that companies with plots below the 45-degree line in the scatter plot may be judged as companies with a high probability of having intangible assets not recorded on their balance sheets.
