关键词: 宏观经济指标, 新三板企业, 转板, 退市, 机器学习

Abstract: The low listing threshold, limited liquidity and relaxed regulatory environment of the National Equities Exchange and Quotations (NEEQ) make investors face the dilemma of quality judgment when choosing enterprises. The existing research on financial crisis early warning of NEEQ enterprises can help investors avoid high-risk enterprises. But it can’t screen out high-quality enterprises with high growth potential. In addition, the existing research often focuses on the internal operating conditions of enterprises, but ignores the impact of macroeconomic factors on enterprise development. Therefore, this study will expand the research perspective from the financial crisis prediction of the NEEQ enterprise to the multi-classification prediction of the transfer, maintenance and delisting of NEEQ, consider the influence of macroeconomic factors on the future development of the NEEQ enterprise, and establish an effective forecasting system of the transfer and delisting of NEEQ enterprises by integrating macroeconomic indicators into enterprise micro-business ones. The purpose of this study is to help investors find high-quality enterprises, avoid investment risks and realize investment returns, so as to promote the rapid development of NEEQ enterprises, realize the goal of “cultivating excellence” of NEEQ and promote the sustainable development of national economy and industry.
   To simultaneously predict the three development states of NEEQ enterprises including the transfer, maintenance and delisting of NEEQ, this study considers the influence of macroeconomic factors on the development of enterprises, selects macroeconomic indicators and constructs the prediction indicator system of NEEQ enterprises’ transfer and delisting with enterprise micro-business indicators. On the basis, a forecasting model is constructed to verify the influence of macroeconomic indicators on the transfer and delisting of NEEQ enterprises. Firstly, according to macroeconomic theory and related research, this study selects macroeconomic indicators such as GDP growth rate, savings rate and consumer confidence index, and integrates them into enterprise micro-business indicators to form a prediction index system of the transfer and delisting of NEEQ enterprises, which provides index system support for the establishment of machine learning classification model. Secondly, the prediction of the transfer and delisting of NEEQ enterprises is transformed into a three-classification problem of machine learning, and the prediction model of the transfer and delisting of NEEQ enterprises is constructed by using OVO decomposition strategy, random forest, GBDT, XGBoost and LightGBM algorithms and Bayesian optimization algorithm. Finally, based on the empirical study of 9334 enterprises that were transferred, kept and delisted from 2020 to 2022, this study analyzes the impact of macroeconomic indicators on the transfer and delisting of the NEEQ enterprises, and provides relevant management suggestions for investors.
   The research results show that: (1)Based on the micro-business indicators of enterprises, the introduction of macroeconomic indicators can significantly improve the prediction effect of the transfer and delisting model of NEEQ enterprises, and the Accuracy, Precision, Recall, F1-Score, Kappa and G-means are increased by 12.15%, 13.18%, 14.32%, 17.22%, 32.16% and 17.89%, respectively. Among the four prediction models constructed in this paper, XGBoost has the best prediction performance, and its Accuracy, Precision, Recall, F1-Score, Kappa and G-means reach 93.39%, 86.22%, 85.28%, 85.62%, 84.34% and 85.40%, respectively.(2)According to the ranking of importance, four of the top five indicators belong to macroeconomic indicators, which are ranked first, second, fourth and fifth, respectively, which verifies the importance of macroeconomic indicators. (3)The most important macroeconomic indicators are the PMI, entrepreneur confidence index, GDP growth rate and money multiplier M2. Therefore, when analyzing the development prospects of enterprises, investors should not only pay attention to the micro-business conditions such as registered capital, scale and profitability of enterprises, but also analyze the current macroeconomic environment, such as PMI, entrepreneur confidence index, GDP growth rate and money multiplier M2, so as to better understand the external environment and internal conditions faced by enterprises and make better investment choices.
   This study extends the research perspective from the financial crisis prediction of NEEQ enterprises to the multi-classification prediction of the transfer, maintenance and delisting of NEEQ enterprises, and brings the macroeconomic indicators into the index system, which not only provides a new research perspective for related research, but also provides investors with an important tool to identify potential risks and discover investment opportunities.

Key words: macroeconomic indicators, NEEQ enterprises, transfer, delisting, machine learning