Machine Learning and Zombie Firms Classification

We investigate whether the machine learning technique helps to identify zombie firms. We also analyze the differences in zombie indicators proposed by previous research.revious studies successfully classified firms as zombies by focusing on whether they receive subsidized credits. However, when the policy interest rate is low, it becomes more challenging to identify zombies, because low-interest payments by firms can be caused by lenders' support to zombies and by low policy interest rates. According to our machine learning approach, we show that we can predict zombie firms from financial information that is publicly available even when the policy interest rate is low. We also find that the financial accounts important for predicting zombie firms differ for every zombie indicator, suggesting that these indicators reflect different aspects of firms' status.