Data Analysis Research Of Continuous Auditing Technology

With the development of computer information technology,the trend of big data became more obvious.Audit data source expanded from traditional business data within the enterprise to more possible data sources.Moreover,the speed of information transmission accelerated and further promoted higher real-time information demand.Traditional audit mode showed great inadaptation to the change of regulatory environment and requirements.How to dig up the effective information from large amounts of data and give timely feedback to the management so as to improve the efficiency and effect of corporate governance has become the main research direction of modern internal audit.The concept of continuous audit provides new strategy for internal audit.First,this paper elaborated the basic theory,development status and application of continuous auditing,focusing on the existing typical practice project which can provide referential ideas to the follow-up construction of continuous auditing system framework of X bank.There were several typical practice projects which adopted different logical structure and realization methods,but all of them mentioned in this paper achieved excellent performance and gave important insights to later research.They are:SAPSECURE?CAMAP?Bagheera-S?pilot project of HSP?pilot project of a south American bank and pilot project of CMCC.We summarized these typical projects and carefully studied the implementation steps of them.Also,we discussed the concept development from continuous auditing to continuous monitoring and explored the nuances of the two concepts.Then we discussed the analysis technologies of current commercial bank loan risk by comparing the advantages and disadvantages of them.New audit technologies emerge endlessly,for example,the discriminant analysis,logistic regression,decision tree,artificial neural network,naive bayesian model and support vector machine.Many scholars tried to apply these models to loan classification analysis.Much of them have proved to achieve very good performance.Also,a lot of scholars compared the accuracy of these models to choose better classification model in predicting credit scoring.It is generally believed that logistic regression?artificial neural network and cluster analysis can achieve relatively high classification accuracy.Then we discussed the necessity and feasibility of the implementation of continuous auditing in X bank based on its current audit system,situation and audit problems.The legal compliance department of X bank performed asset business audit once a year.Due to the large number of business,the auditors just took random sampling inspections of single business most times.There were a lot of problems in X bank's internal audit,for example,audit could not keep up with the pace of business data,auditors did not make full use of the past business data and low efficiency.Taking into account auditing thought and computer technology,we put forward the continuous auditing system framework for X bank which mainly includes the application system and the support system.The application system included five modules:project management module,data acquisition module,data processing module,data analysis module and electronic alarm module.Each of them had their own unique functions.The five modules could almost cover the whole process of continuous auditing in order to realize the corresponding audit purpose,and they were accord with the actual situation of X bank currently.In general,the support system contains communication network system,data management system and document management system.It is performed as an information integration platform thus to support the realization of the function of the application system.Under constraints of this framework,we put focus on the data analysis module.In selecting the model,we mainly consider the following factors:First,prediction accuracy of the model.Among the commonly used models at present on international loan data,logistic regression,artificial neural network,naive bayesian model,support vector machine and clustering analysis showed relatively high prediction accuracy.Second,characteristics of our data.From the current situation of X bank,neither of the personal credit data or the small micro enterprise credit data has high data dimensions,so high-dimensional data analysis techniques(artificial neural network and support vector machine)had no advantages.Also,loan data showed weak non-correlation which could not meet the assumption of naive bayesian model.Last,mathematical model.In a continuousauditing system,we hope to get an accurate mathematical model which can be embedded in the data analysis module in order to reach the purpose of real-time detection of business data.Based on the above considerations,we chose a supervised learning algorithm(logistic regression)and an unsupervised learning algorithm(clustering analysis)to build up the analytic models using personal credit data and small micro enterprise credit data extracted from X bank.We set up six independent variables for personal credit data and nine independent variables for small micro enterprise credit data.At last,we ran clustering analysis model using SPSS software and qualitative response model using Eviews7.0 software.We documented these results and verified the validity of analysis.Also,we compared the prediction accuracy under different situations and found that both of the models could detect most default cases.Both of them showed very good prediction accuracy.By contrast,the qualitative response model performed better than the clustering analysis model.Thus,the data analysis model we have constructed in this paper is feasible,it can serve as the reference model for X bank loan business.Therefore,we suggest X bank to adopt the two qualitative response models as the model embedded in the data analytic module of their continuous auditing system.The significance of this paper lies in two aspects.In theory,continuous auditing method is not only an innovation for internal audit means,but also a leap in audit mode.It will be beneficial to build more scientific internal audit theoretical system.In practice,this paper set up two models for the data analysis module in continuous auditing system which clearly showed the whole process of data modeling using personal credit data and small micro enterprise credit data.The work of this paper provides convenience for future construction of continuous auditing system of X bank and other similar commercial banks.