被引排行
Is Risk Management with Big Dada Effective?——Comparison and Analysis Based on Statistics Score Card and Machine Learning Model
LIU Zhi-hui;HUANG Zhi-gang;XIE He-liang;School of Ecomomics and Management,Fuzhou University;School of Finance,Fujian Business University;School of Statistics and Mathematics,Central University of Finance and Economics;With the great development of financial technology,machine learning has been deeply applied in the field of financial risk management.As one of the most widely used risk assessment models,credit scorecard model has the limitation on comprehensive analysis of high dimensional,complex and nonlinear personal credit data in the big data era.Starting from the actual situation of the development of Internet finance in China,it presents an innovative risk management model of Internet finance based on XGboost machine learning algorithm,and compares with the traditional statistical scorecard model.Also the method of how to transform the predicted results of machine learning model into credit scoring is given.The results show that machine learning model can predict individual credit risk better and a more effective risk management system could be built based on it.
Research on Predictive Modeling on Stock Market Index Based on CEEMDAN-LSTM
HE Yi-yue;LI Ping;HAN Jin-bo;In order to meet the demand of high-precision forecast of stock market index in the field of active quantitative investment,complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) is introduced into predictive modeling on stock market index.Combing CEEMDAN with Long Short-Term Memory(LSTM),which can model long-term dependencies of complex time series efficiently,an integrated prediction method CEEMDAN-LSTM for stock market index is proposed following "decomposition-reassembly-prediction-integration" process.In the method,CEEMDAN is applied to decompose the index and reconstruct its high frequency component,low frequency component and trend items.Then,the LSTM prediction model is established for each component respectively,and the predicted values of each component are furtherly integrated as the overall predicted index value.Finally,comparative experiments are conducted and the prevalent machine learning modeling methods of financial time series,with five representative stock market indexes including CSI 300 as test data.The experimental results prove that CEEMDAN-LSTM has lower prediction error and lag,and its prediction performance is significantly better than that of the existing modeling methods.
Constructing the China Customer Satisfaction Indices System
JIN Yong-jin, WANG Hua (School of Statistics, Renmin University of China, Beijing 100872)It is necessary for China to construct Customer Satisfaction Indices system in national level as well as in industrial level. The paper probes into forming several rules constructing the Chinese Customer (Satisfaction) Indices (CCSI) system, determining the six levels of CCSI and compounding methods in every (level.) Finally, relevant weighting compounding variables have been determined.
Issues in Spatial Panel Data Model Specification
JI Min-he1,WU Zhan-yun2,JIANG Lei1(1.Key Laboratory of GIScience,Ministry of Education,ECNU,Shanghai 200062,China;2.Urban Planning and Architectural Design Institute,Fudan University,Shanghai 200043,China)Combining spatial econometrics and panel data modeling,the spatial panel data model considers both spatiotemporal features and spatial effects,rendering it a hot topic in the mainstream econometrics.Therefore,the specification,estimation and diagnostics of spatial panel data model is more complicated,misspecification always arise in empirical study.Based on the newest theory of the spatial panel model,this paper discussed the problem of specification in empirical study,including the choice of spatial lag model and spatial error model,the choice of random effect and fixed effect,and the goodness fit of the model,which provide a stimulus and resource to theoretical and applied researchers alike to aid in pursing these directions in the future.
Intrinsic Mechanisms and Spatial Spillover Effects of the Digital Economy Impacting Common Wealth
WANG Jun;LUO Xi;As an important goal of the development of socialism with Chinese characteristics in the new era, the development of common prosperity is related to people's expectations for a better life.As a ‘new engine' of economic development, the impact of digital economy on common prosperity remains to be demonstrated.In this paper, the impact, mechanism and spatial spillover effect of digital economy on common prosperity are analyzed from a theoretical perspective, and the empirical tests are conduct based on panel data of China's provinces(municipalities and autonomous regions).The results show that during the study period, the development of digital economy has significantly improved the development level of common prosperity, and has a non-linear characteristic of increasing positive ‘marginal effect' on the promotion of common prosperity, and has passed the robustness test.The analysis of intermediary mechanism shows that the digital economy promotes the development of common prosperity by acting on the upgrading of industrial structure and narrowing the ‘digital divide'.In addition, through the analysis of spatial Dubin model, it is found that the positive impact of digital economy on common prosperity has spatial spillover characteristics.The research is helpful to explain the motivation, mechanism and effect of China's common prosperity development in the era of digital economy, and provide reference for China's common prosperity development and construction.
Enterprise Digital Transformation and Capital Allocation Efficiency
LI Qin-yang;ZHI Jia;LIU Xiang-qiang;In the era of digital economy, digital transformation has become an important path for the innovation-driven development of enterprises.The study on the internal logical relationship between digital transformation and enterprise capital allocation efficiency has important theoretical and practical significance for promoting firms to carry out digital transformation and achieve efficiency reform.Therefore, taking Chinese A-share listed firms from 2008 to 2020 as samples, using Python crawler technology and text analysis method to construct indicators to measure the degree of enterprise digital transformation, and empirically testing the relationship between enterprise digital transformation and capital allocation efficiency.It is found that digital transformation of enterprises can significantly improve their capital allocation efficiency.The above relationship is established after several robustness tests.The heterogeneity test shows that there is heterogeneity in the impact of digital transformation on capital allocation efficiency of enterprises.Among enterprises with higher industry competition, enterprises with serious agency conflicts, high-tech enterprises and smaller scale enterprises, digital transformation plays a more significant role in improving the efficiency of enterprise capital allocation.The extended research shows that the positive impact of enterprise digital transformation on capital allocation efficiency will further promote the growth of firm value.Therefore, enterprises should fully grasp the development opportunities of digital transformation, actively implement the digital transformation strategy, and create a good information environment, so that the digital transformation can deeply integrate with the enterprise development orientation, promote the improvement of the enterprise operation management level, and then promote the enterprise to achieve the dual goals of efficiency transformation and value growth.The government should focus on building and improving the incentive policies and support mechanisms for enterprises' digital transformation to help enterprises implement the digital transformation strategy.
Research on Supply Chain Management of Working Capital Evaluation Index System:Based on Supply Chain of the Coal Industry
XU Wei1,2(1.School of Management Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China; 2.School of Business,Xi’an University of Finance and Economics,Xi’an 710100,China)Establishing a set of perfect supply chain operation capital evaluation index system is the infrastructure of supply chain enterprises’ fund management,and it’s also a way to assess financial management level of supply chain.From the view of characteristic of supply chain management,it has put forward the working capital evaluation index system by using the analytic hierarchy process,then striving the point and face,in order to provide theory and method which support for working capital management evaluation system of supply chain enterprises.
Discussion of the Same Trending Methods of Indices in Principal Component Analysis and Factor Analysis
CHEN Jun-cai (Department of Statistics, Guangdong College of Business, Guangzhou, 510320 Guangdong)Sample principal component analysis and factor analysis has been one of the main methods of composite analysis, the communalities of variables is a key step of using these methods. In this paper, we summed up the methods of communalities of variables in the principal component analysis and factor analysis, discussed the effects of these methods have on composite analysis and pointed out the conditions of using them.
Logistic Model Based on Benford's Law and Its Application in Fraud Detection
YANG Gui-jun;ZHOU Ya-meng;SUN Ling-li;SHI Yu-hui;School of Statistics,Tianjin University of Finance and Economics;Both Benford's law to assess the quality of financial samples and Logistic model to have high classification accuracy rate are very often useful approaches for fraud detection.A new method of detecting fraud is proposed by combining Logistic model with Benford's law.Benford's law is used to screen financial indies with low data quality,the Benford factors of potential anomalous samples are constructed,and a Logistic model containing the Benford factors is established.Based on the financial data from China's listed companies,the analyzed results show that the new model tends to have higher classification accuracy.And the Logistic model with the Benford factors often has more classification accuracy for financial fraud samples than the general Logistic model.
The Theory and Methods of Data Preparation: An Overview
CHENG Kai-ming(School of Statistics and Mathematics,Zhejiang Gongshang University, Hangzhou 310018,China)In order to improve the quality of data for analyzing,data must be prepared.Data preparation can be decomposed to four steps such as data examination,data cleaning,data transformation and data validation.The methods of data preparation include descriptive and exploratory analysis,missing data analysis,outlier processing,transformation techniques,reliability and validity analysis,and national economic data diagnosis.Data preparation can be operated by software and some possible problems must be noticed.