基于轻量级卷积神经网络的科技型中小企业估值研究

Research on Valuation of Technological Small and Medium-sized Enterprises Based on Lightweight Convolutional Neural Network

摘要:

对科技型中小企业进行科学合理估值有利于其募集资金从而健康发展，但传统估值方法的准确度和适用性均有待考量。鉴于此，利用文献分析与关键词提取方法，初步从相关文献中提取科技型中小企业估值指标，并通过德尔菲法筛选得到能体现科技型中小企业核心能力的关键指标；进一步设计轻量级卷积神经网络（TecNet）结构，基于大数据训练构建适用于科技型中小企业的估值模型；最后以10家样本企业为例，对其2017－2021年度预测值与真实值相对误差进行拟合分析以及敏感度分析，以验证估值模型的有效性。结果表明：模型预测的拟合曲线与真值重合度较高，平均相对误差为3.43%，且90%的样本的相对误差在10%以内，说明TecNet模型具有科学有效性；将关键指标提升5%时，硕士及以上比例、研发投入占营业收入比例以及专利数对估值结果浮动程度最高，分别为16.51%、16.10%和11.87%，表明科技型中小企业价值对这3项指标最为敏感。因此，科技型中小企业发展应着眼于其关键指标，通过优化员工学历层次、加大研发投入等措施提高智力资本，并持续加强技术创新。

Abstract:

Scientific and reasonable valuation of technology-based small and medium-sized enterprises (SMEs) is conducive to their healthy development by facilitating capital raising. However, the accuracy and applicability of traditional valuation methods are subject to considerations. In view of this, utilizing literature analysis and keyword extraction methods, this study initially extracts valuation indicators for technology-based SMEs from relevant literature. Through the Delphi method, key indicators that can reflect the core capabilities of technology-based SMEs are selected. Furthermore, a lightweight convolutional neural network (TecNet) architecture is designed. Based on big data training, a valuation model suitable for technology-based SMEs is constructed. Finally, using a sample of 10 enterprises, the relative errors between predicted and actual values for the years 2017-2021 are subjected to fitting analysis and sensitivity analysis to validate the effectiveness of the valuation model. The results show that the fitted curve of the model's predictions closely matches the true values, with an average relative error of 3.43%. Additionally, 90% of the sample's relative errors are within 10%, indicating the scientific validity of the TecNet model. When key indicators are increased by 5%, the proportions of master's and above degrees, research and development investment to operating income ratio, and the number of patents have the highest impact on valuation results, at 16.51%, 16.10%, and 11.87% respectively. This demonstrates that the value of technology-based SMEs is most sensitive to these three indicators. Therefore, the development of technology-based SMEs should focus on their key indicators, optimize employee education levels, increase research and development investment, and continue to enhance technological innovation.