上市公司财务数据造假的情况会给股民及证券市场带来巨大的经济损失。因此投资者若能在投资前识别出数据造假的公司进行规避投资，就能有效降低投资风险，保障资金安全。
数据集中的所有上市公司来自 19 个行业，其中制造业的公司是最多的，有 2667 家。其他行业的上市公司数量分布相对均匀，且数量较少。因此我们就将整个数据分成制造业和其他行业，在此基础上训练的模型效果会更好。
首先，财务风险评估指标体系的构建： SMOTE 采样数据训练机器学习算法去求对公司财务数据造假有较大影响的特征，基于投票打分机制选出排名前30的特征作为不同类型上市公司财务数据评估的关键指标。其次，财务造假预测模型的搭建：本文用深度学习模型代替了传统的机器学习模型，以多层感知机，多层残差网络，Cross 网络作为子网络构建了(DCRN)网络模型，并进行 Bagging 集成提高模型的泛化能力。最终预测效果好，稳定性好，给出了下一年有财务造假风险的公司名单。
本文使用了机器学习算法挑选特征，融合多种深度学习算法模型建立了 Baggin+DCRN 集成学习模型，具有较高的参考价值和实际意义。
Fraudulent financial data of listed companies can cause huge financial losses to shareholders and the securities market. Therefore, investors can effectively reduce investment risks and protect their capital if they can identify companies with falsified data for investment avoidance before investing.
All listed companies in the data set come from 19 industries, with the manufacturing industry having the largest number of companies, with 2667. The number of listed companies in other industries is relatively evenly distributed and the number is small. Therefore, we divided the whole data into manufacturing and other industries, and the model trained on this basis will be more effective.
Firstly, the construction of financial risk assessment index system: this paper uses SMOTE sampling data to train the classical machine learning algorithm to find the features that have a greater impact on the company's financial data falsification, and selects the top 30 features based on the voting scoring mechanism as the key indicators for financial data assessment of different types of listed companies. Second, the construction of the financial falsification prediction model: this paper replaces the traditional machine learning model with a deep learning model, constructs a (DCRN) network model with a multi-layer perceptron, a multi-layer residual network, and a Cross network as sub-networks, and carries out Bagging integration to improve the generalization ability of the model. The final prediction is good and stable, giving a list of companies at risk of financial fraud in the following year.
This paper uses a machine learning algorithm to select features and fuses multiple deep learning algorithm models to build a Baggin+DCRN integrated learning model, which is of high reference value and practical significance.
[1]韩世强. 上市公司财务造假问题研究综述[J]. 合作经济与科技, 2023, 10(6): 148-150.
[2]李文. 上市公司财务舞弊的防范措施——A公司财务造假案例分析[J]. 经济研究导刊, 2022, 10(36): 131-134.
[3]王圣洁. 基于XGBoost算法的上市公司财务造假识别问题研究[D]. 青岛大学, 2022.
[4]CEO tenure and audit pricing[J]. Santanu Mitra; Hakjoon Song; Sang Mook Lee; Shin Hyoung Kwon.Review of Quantitative Finance and Accounting, 2020, 9(2): 22-25.
[5]Natter M. Conditional Market Segmentation by Neural Networks[J]. Journal of Retailing & Consumer Services, 1999, 6(4): 237-248.
[6]Trapat S. An Investigation of Thai Listed Firms’Financial Distress Using Macro and Variables[J]. Multinational Finance Joumal, 1999, 8(6): 31-39.
[7]Martin, D.Early Waring of Bank Failures: A Logit Regression Approach[J]. Journal of Banking and Finance, 1977, 7(9): 939-945.
[8]赵惠芳, 王晓, 徐晟. 基于Logistic回归模型的上市公司财务危机预警分析[J]. 金融探讨, 2011, 10(1): 97-100．
[9] Krizhevsky A, Sutskever I, Hinton G E. Image Net classification with Deep Convolutional Neural Networks[J]. Advances in Neural Information Processing Systems, 2012, 25(2): 103-120.
[10]Qin Xu, Yong Xiao, Dongyue Wang, Bin Luo. CSA-MSO3DCNN: Multiscale Octave 3D CNN with Channel and Spatial Attention for Hyperspectral Image Classification[J]. Remote Sensing, 2020, 12(1): 37-40.
[11]曹兴. 基于深度学习网络的企业财务风险预警研究[D]. 中国地质大学(北京), 2016.
[12]宋歌, 马涛.基于深度学习的上市公司财务风险预警模型研究[J]. 价值工程, 2019, 38(1): 53-56.
[13]邹新颖, 涂光华.基于信用风险及融资约束模型的财务预警指标构建[J].统计与决策, 2015, 7(14):45-48.
[14]Altman E I. Financial ratios, discriminant analysis and the prediction of corporate
bankruptcy[J]. The journal of finance, 1968, 23(4): 589-609.
[15]Xinzhong B, Guangshuo H. A Study of Listed Companies'Financial Distress Prediction Using Rough Set Conditional Entropy Method, International Conference on Information Management[J]. Innovation Management and Industrial Engineering, 2010, 9(1):460-463.
[16]Fitz Patrick P.J.A Comparison of Ratios of Successful Industrial Enterprises with
Those of Failed Firms”, Certified Public Accountant[J]. Journal of Accounting
Research, 1932, 9(8):598-605.
[17]肖珉. 我国企业集团上市公司财务预警与信用风险评估研究[D]. 电子科技大学, 2012.
[18]Kłosok M, Chlebus M.Towards better understanding of complex machine learning models using Explainable Artificial Intelligence（XAI）-case of Credit Scoring modelling[R]. 2020.
[19]Cui D. Financial Credit Risk Warning Based on Big Data Analysis[J]. Metallurgical&Mining Industry, 2015, 10(6): 97-99.
[20]李成刚, 贾鸿业, 赵光辉, 付红.基于信息披露文本的上市公司信用风险预警—来自中文年报管理层讨论与分析的经验证据[J].中国管理科学, 2021, 9(8):1-14.
[21]吴革, 叶陈刚.财务报告舞弊的特征指标研究:来自A股上市公司的经验数据[J].审计研究.2008, 6(1):34-41.
[22]万希宁, 王艳.基于非财务指标的企业财务危机模糊预警模型研究[J].管理学报, 2007, 2 (4):195-200.
[23]赵春.基于数据挖掘技术的财务风险分析与预警研究[D].北京化工大学, 2012.
[24]春光、程晓娟.企业财务危机预警指标筛选探究[J].当代经济管理.2010, 9(8):69-71.
[25]田青, 王宝帅.基于数据挖掘技术的上市公司预警模型的比较研究[J].中国管理科
学, 2010, 7 (18):289-292,
[26]郭毅夫, 权思勇.基于神经网络的创新型企业财务危机预警研究[J].统计与决策, 2013, 8(4):170-172.
[27]Tokushige S, Yadohisa H, Inada K. Crisp and fuzzy K-means clustering algorithms for multivariate functional data[J]. Computational Statistics, 2007, 22(1): 1-16.
[28]Shen H, Cui J, Zhou Z, Min H.BP-Netural Network Model for Financial Risk Warning in
Medicine Listed Company[C]. Fourth International Joint Conference on Computational
Sciences and Optimization(CSO 2011), 2011, 8(4): 15-19.
[29] SEVIM C, OZTEKIN A, BALI O, et al. Developing an early warning system to predict currency crises[J]. European Jour⁃nal of Operational Research, 2014, 237(3):1095-1104.
[30]De Servigny, Renault. Measuringand Managing Credit Risk[M]. New York: Mc-Graw-Hill, 2004.
[31]潘彬, 凌飞. 引入违约距离的上市公司财务危机预警应用[J]. 系统工程, 2012, 3(30): 45-51.
[32]尹杞月. 信用价值与中小企业融资可持续发展[J]. 商业研究, 2013, 8(5):73-80.
[33]叶晓枫, 鲁亚会.基于随机森林融合朴素贝叶斯的信用评估模型[J].数学的实践与认识, 2017, 2(2):68-73.
[34]Macqueen JB. Methods for Classification and Analysis of Multivariate Observations[J]. Proc. Symp. Math. Statist and Probability, 1967, 8(5):67-80.
[35]Cios KJ, Pedrycz W, Swiniarski RW. Data Mining and Knowledge Discovery[M]. Springer US, 1998.
[36]曹彤, 郭亚军.基于神经网络模型的上市公司财务风险预警研究——来自山东省制造业
数据[J].财会通讯, 2014, 13(9):89-92.
[37]杨建辉, 李龙.基于SVR的期权价格预测模型[J].系统工程理论与实践, 2011,
31(5):848-854.
[38]Tarun K.Sen, Parviz Ghandforoush, Charles T.Stivason.Improving prediction of neural
networks:a study of tow financial prediction tasks[J].Journal of Applied Mathematics and
Decision Sciences, 2004, 8(4):219-233
[39]Min, J.H., Lee, Y.C.Bankruptcy Prediction Using Support Vector Machine with Optimal
Choice of Kernel Function Parameters[J].Expert Systems with Applications, 2005, 28(1):128–134.
[40]杨杰群.基于深度学习之股指期货交易[D].中国科学技术大学, 2015.
[41]Dahl G E, Yu D, Deng L, et al.Context-Dependent Pre-Trained Deep Neural Networks for
Large-Vocabulary Speech Recognition[J].IEEE Transactions on Audio Speech&Language
Processing, 2011, 20(1):30-42.
[42]丁德臣.混合HOGA-SVM财务风险预警模型实证研究[J].管理工程学报, 2011,
25(2):37-44.
[43]Krizhevsky A , Sutskever I , Hinton G . ImageNet Classification with Deep
Convolutional Neural Networks[J]. Advances in neural information processing
systems, 2012, 25(2):102-130.
[44]Lu Y, Roychowdhury V. Parallel randomized sampling for support vector machine (SVM)
and support vector regression (SVR)[J]. Knowledge & Information Systems, 2008, 14(2):233-247.
[45] Rana R. Gated Recurrent Unit (GRU) for Emotion Classification from Noisy Speech[J].
2016, 7(1):90-94.
