Title: The application of deep learning to predict corporate growth
Authors: Sumitaka Ushio; Nobuhisa Yamamoto
Addresses: Chuo University, 742-1, Higashinakano, Hachioji City, Tokyo, Japan ' Sharp Corporation, 1 Takumi-cho, Sakai-ku, Sakai City, Osaka, Japan
Abstract: This study examines the use of deep learning to predict corporate growth. An algorithmic model is constructed to identify growing (as well as non-growing) companies based on a snapshot (single year) of financial data without a time series. The binary classification model predicts whether sales will increase in the following year for 353 retail companies in the Tokyo Stock Exchange 33 sector category in Japan, by utilising all available items in their balance sheets and profit/loss statements (308 numerical values) as well as the size of the companies. As a result, the model achieves 74.79% classification accuracy. The area under the curve (AUC) of the model is 0.75, which shows moderate accuracy of prediction regardless of its cut-off point. This study also debates the methodological significance of applying deep learning to accounting research in comparison with traditional (frequentism) statistics.
Keywords: deep learning; growth prediction; accounting research; artificial intelligence; AI; algorithmic modelling.
DOI: 10.1504/IJEA.2020.10032541
International Journal of Economics and Accounting, 2021 Vol.10 No.2, pp.248 - 263
Received: 26 Oct 2019
Accepted: 09 Jun 2020
Published online: 24 May 2021 *