通過隱含狄利克雷分布演算法預測盈餘管理

Abstract

在已知Brown et al. (2020)的方法可以有效預測財務舞弊案的情況下，我們探討該方法是否也可以用於預測盈餘管理。該方法主要是把一種稱爲隱含狄利克雷分布演算法 (Latent Dirichlet Allocation) 的機器學習方法應用於日漸複雜至人類難以理解的公司年報上。盈餘管理預測能讓公司真正的財務狀況更清楚地展示出來。這可以提供有效的資訊于投資者、債權人、審計師、立法委員等等以便他們可以做出更好的決策。此爲，利用該方式來同時預測盈餘管理以及財務舞弊可以降低成本及時間。

We examine whether the Brown et al. (2020) method that can predict financial fraud using annual report filings is also able to predict earnings management. This model applies Latent Dirichlet Allocation (LDA) an advanced machine learning tool to the annual reports’ qualitative disclosures that have become incomprehensible to humans over time. Predicting earnings management paints a clearer picture of the firm’s financial position. This will allow the investors, creditors, auditors, legislators, and other stakeholders to make higher-quality decisions. Besides, using the proposed approach in the context of both abnormal accruals and financial misreport can save time and money.