School of Economics and Management, Nanjing University of Science and Technology, Nanjing City 210094, China.
School of Economics and Management, Chuzhou University, Chuzhou City 239000, China.
Comput Intell Neurosci. 2022 May 18;2022:7297769. doi: 10.1155/2022/7297769. eCollection 2022.
At present, there are still some problems in the document management of enterprise innovation projects, such as non-standard management, lagging update, chaotic content, insufficient information, and insufficient application. There is still a lack of effective methods to evaluate the financing ability of enterprises. To solve the above problems, high technology expertise (HNTE) is taken as the research objects. Firstly, the relationship between social audit and enterprise technological innovation is analyzed, and on this basis, combined with natural language processing (NLP), an extraction method of project document information is proposed. Secondly, the evaluation index system of enterprise financing ability is constructed based on Back Propagation Neural Network (BPNN), and the technology innovation audit system of HNTEs. Finally, combined with the actual content, the proposed document audit method is evaluated. The results show that: the average accuracy rate of the NLP-based innovation project document audit method is 91.36%, the average recall rate is 96.34%, and the average F statistical value is 95.34%. Among them, the recall rate and F statistical value are about 2.3% and 1.4% higher than manual processing, respectively. The recall rate and F value are obviously better than those of manual processing methods, and the processing time of single document based on NLP is only 87.5 s. The processing time is nearly 50 times lower than that of manual processing, which greatly improves the processing efficiency of document information. The corresponding test results of each index selected based on the evaluation of enterprise financing ability are all below 0.1, which meets the requirements of consistency. The evaluation results of BPNN model on enterprise financing ability are highly consistent with the target value, and the prediction error is controlled within 0.02, which can provide more accurate prediction results. This research obtains a more accurate prediction model of enterprise financing ability evaluation, which provides technical support for social auditing and the innovation and development of enterprise technology, and provides a feasible route for the development of BPNN in the financial field.
目前，企业创新项目文档管理仍存在规范性差、更新滞后、内容混乱、信息量不足、应用不足等问题。企业融资能力的评估方法还比较缺乏。为了解决上述问题，以高新技术企业（HNTE）为研究对象。首先，分析了社会审计与企业技术创新的关系，在此基础上，结合自然语言处理（NLP），提出了一种项目文档信息提取方法。其次，基于反向传播神经网络（BPNN）构建了企业融资能力评价指标体系，并构建了 HNTE 技术创新审计体系。最后，结合实际内容，对提出的文档审计方法进行了评价。结果表明：基于 NLP 的创新项目文档审计方法的平均准确率为 91.36%，平均召回率为 96.34%，平均 F 统计值为 95.34%。其中，召回率和 F 统计值分别比人工处理高约 2.3%和 1.4%。召回率和 F 值明显优于人工处理方法，基于 NLP 的单个文档的处理时间仅为 87.5 s。处理时间比人工处理低近 50 倍，大大提高了文档信息的处理效率。基于企业融资能力评价选择的各指标的对应测试结果均低于 0.1，符合一致性要求。BPNN 模型对企业融资能力的评价结果与目标值高度一致，预测误差控制在 0.02 以内，可以提供更准确的预测结果。本研究获得了更准确的企业融资能力评价预测模型，为社会审计和企业技术创新发展提供了技术支持，为 BPNN 在金融领域的发展提供了可行途径。
