百年中大·创新创业与经济高质量发展大讲坛第5期暨相山大讲堂第15期 | 柯滨

 

 

　　讲座主题

　　The Value of Historical Financial Statements in Predicting the Innovation Potential of Technology-Based SMEs: A Machine Learning Approach

　　讲座嘉宾

　　柯滨

　　讲座时间

　　2024年5月28日（周二）10:00-11:30

　　讲座地点

　　中山大学（深圳校区）文学园1栋202会议室

　　

　　

　　讲座嘉宾介绍

　　柯滨（Bin Ke）教授，美国密歇根州立大学博士，曾任教于宾夕法尼亚州立大学，新加坡南洋理工大学，现任新加坡国立大学商学院会计系教授、教务长。2010年获聘国家级重大人才工程。曾任北美华人会计教授会（CAPANA）会长。兼任The Accounting Review、Journal of American Taxation Association、The International Journal of Accounting编委。柯滨教授先后在The Accounting Review、Journal of Accounting and Economics、Journal of Accounting Research、Review of Accounting Studies和Contemporary Accounting Research上发表多篇学术论文。柯滨教授擅长从经济学的角度研究会计信息的生成与运用过程，主要研究领域包括盈余管理、内幕交易、机构投资者与财务分析师行为等，近期主要关注新兴市场（尤其是中国）的财务报告、管理层激励和投资者保护问题。

　　

　　

　　讲座摘要

　　Utilizing a proprietary dataset of grant applications from technology-based SMEs to China's Innovation Fund for Small Technology-based Firms (Innofund), this study evaluates the predictive utility of historical financial statements for assessing SMEs' innovation potential. Our findings demonstrate that an advanced XGBoost model, using only financial statements, predicts innovation potential more effectively than an XGBoost model based on human expert evaluation scores only. Additionally, merging financial data with expert scores does not improve prediction accuracy, indicating no incremental value from expert discretion. Analysis of SHAP values highlights distinct evaluation criteria differences between human experts and the XGBoost model. Our results demonstrate the promise of leveraging advanced machine learning with historical financial data in aiding government grant agencies to identify SMEs with high innovation potential.