The Research And Implementation Of Intelligent Financial Audit System Based On Data Mining

It is difficult for the traditional audit method to find out various financial problems hiding in massive amounts of data.Therefore,this paper combines data mining technology,expert diagnosis theory and financial audit theory,and uses the interdisciplinary thinking mode to develop an intelligent audit system that can deal with a variety of data types and find audit clues automatically.The main research contents of the paper are as follows:(1)Research on Auditing Analysis Model of Financial Statements.The paper innovatively applies the expert diagnosis theory to the field of financial auditing,and regards the key audit objects found in the overall analysis model as suspected fault points,and combines the key analysis model and the individual analysis model to conduct deep and three-dimensional inference analysis of financial data to explore abnormal clues,and use knowledge pool to store audit knowledge rules,audit recommendations and audit models.the utilization rate of the model and the work efficiency of auditors are improved.(2)Research on intelligent analysis algorithms for accounting voucher.Aiming at the problem that the unit name records in the voucher abstract and accounting subjects are not standardized,this paper firstly designs a voucher content normalization algorithm based on set theory to realize the standardization of vouchers.On this basis,a credential digest clustering algorithm based on word co-occurrence and SOM network is proposed.Text vector representation model was established by calculating the co-occurrence rate between words and using TextRank algorithm to construct co-occurrence graph to obtain co-occurrence phrases,and the improved SOM algorithm was used to realize abstract clustering,and combines the correspondence of accounting subjects to find audit doubts.(3)Research on audit reports intelligent analysis and processing.In order to achieve the classification management of audit problems,a feature fusion short text classification algorithm based on BiLSTM neural network is proposed.The algorithm combines the feature vectors extracted by Word2 vec,TF-IDF and LDA into an input matrix,and inputs it into BiLSTM to utilize the hidden layer.The node obtains the internal relationship of the text,and finally inputs the classified auditing questions,laws and regulations into the database,and visualizes the auditing problem.(4)database log analysis and processing.The log files of the underlying database of financial software are extracted and visualized,and the information such as operation time,operation type and operation frequency are used to determine whether the user's behavior is abnormal and provide audit clues.The validity of the credential abstract clustering algorithm based on word co-occurrence and SOM neural network and the feature fusion short text classification algorithm based on BiLSTM neural network was verified through experiments.On the basis of the above algorithms,an intelligent financial audit system was developed.