SMART FINANCIAL AUDIT SYSTEM

Abstract

This invention relates to a system (1) that enables the detection of suspicious and abnormal transactions among accounting data using artificial intelligence algorithms.

Description

1 TARIFF SMART FINANCIAL AUDIT SYSTEM Technical Area This invention uses artificial intelligence algorithms to identify differences between accounting data. It relates to a system that enables the detection of suspicious and abnormal transactions. Previous Technique Today, machine learning is used to detect abnormal behavior in financial transactions. Systems that identify and analyze contributing factors are used. However, These systems are designed to detect suspicious data in accounting records. from analyzing contributing factors, abnormality detection results due to its interpretability, risk calculation related to accounting transactions and 15 verification is carried out in accordance with GDPR (General Data Protection Regulation) or KVKK (Turkish Personal Data Protection Law). (Such as the Personal Data Protection Law in Türkiye) compliance, taking into account actual and recent transactions, specific from focusing on types of financial transactions and specific data sets, to complex 20 from the use of artificial intelligence algorithms and modeling techniques It does not mention it. Chinese patent CN114493864A, which falls under the known state of the art. The document describes accounting anomaly detection and methods based on big data. 25 from an AI-powered system structured for classification. This discovery is mentioned. This invention involves a fund that includes the finance sector and big data anomalies. It presents a system and method based on a detection system and method. The system includes a data collection unit, which collects data related to basic fund flows. It is used to collect and intelligently import, then create a fund data pool. It is formed. In addition, the data management unit extracts the basic fund 30 from the fund data pool. retrieving, classifying, and organizing data related to the stream, then 2 to convert and clean and create the basic fund database It is used. The evaluation and decision unit uses the fund trading feature data pool. Based on this, it is possible to evaluate fund trading holistically and the relevant fund accounts. It is used to determine whether something is abnormal. Fund characteristics, for example 5 factors such as trade volume, trade time, trade location, trade frequency, and trade method. Feature-based fund identification technology, data statistics methods, and computers. These features are analyzed using ranking matching algorithms. Data trends, ratios, etc. are analyzed to determine the nature of fund trading. to analyze the flow of trade and quickly identify key suspicious accounts and It is used to determine the funds. 10 Brief Description of the Invention The goal of this invention is to use artificial intelligence algorithms to analyze accounting data. a system that enables the detection of suspicious and abnormal transactions between them 15 to accomplish. Detailed Description of the Invention The "Smart Financial Auditing System" 20 was developed to achieve the purpose of this invention. This is shown in the attached figure; Figure 1. Schematic view of the system described in the invention. The parts shown in the figure are individually numbered, and the corresponding numbers are 25. given below: 1. System 2. Database 3. Server 30 3 Using artificial intelligence algorithms, suspicious and suspicious items in accounting data can be identified. The system in question enables the detection of abnormal operations (1); - data related to accounting transactions, artificial intelligence models and models Training data and suspicious transaction reports used in training are kept on record. at least one database (2) and 5 configured to ensure its retention - Collection and preparation of accounting data, identification of suspicious transactions selecting and developing the model to be used, training the model, and verification, real-time suspicious transaction detection using the model, detection reporting of the transactions, taking security measures, and modeling at least one server configured to improve its performance (3) 10 It includes. In the system that is the subject of the invention, the database (2) is in communication with the server (3) (1). It is configured to exchange data with the server (3). In the preferred arrangement of the invention, the general accounting database (2) contains 15 Raw data obtained from sources such as ledgers and trial balances, receivables Additional data collected from accounts and payable records, selected artificial intelligence training datasets to be used in training the models, real-time Monitoring data related to accounting records that were found to be suspicious. keeping records of notifications created and reports prepared for transactions. 20 It is structured to provide this. The server (3) in the system (1) which is the subject of the invention communicates with the database (2) and is configured to access the data in the database (2). Server (3), raw 25 from external servers in the form of general ledger and trial balance sheet. It is configured to enable the collection of data. The server (3) will receive by accessing account and payable records from external servers, additional information can be obtained. It is configured to collect the collected data. The server (3) to ensure cleaning, normalization and completion of missing data It is structured in such a way. The server (3) is structured to 30 the dataset through feature engineering. to ensure enrichment and extraction of necessary properties 4 It is configured. Server (3) is used for anomaly detection. Isolation Unsupervised algorithms such as Forest or One-Class SVM It is configured to enable selection. The server (3) is for classification. Random Forest, Gradient Boosting Machines (GBM) or deep learning models are used. 5 It is configured to provide time series analysis and sequential. Server (3) algorithms such as ARIMA, RNN, LSTM or GRU for models It is structured to enable the evaluation of the selected Server (3). to enable algorithms to be trained on the training dataset is being configured. Server (3), cross-validation and other 10 Evaluating model performance using validation techniques It is configured to provide the server (3) with hyperparameter adjustment. It is structured to enable the model to be improved. Server (3), to ensure the integration of the developed model with existing accounting software is configured. Server (3), automatic model update and 15 to establish a workflow for retraining It is configured. The server (3) is integrated into the real-time data stream of the model. It is configured to ensure that it is monitored and continuously tracked. Server (3), To ensure the setup of alarm and notification systems for detecting suspicious transactions. It is structured in such a way. The server (3) detects 20 suspicious transactions. It is configured to enable reporting. Server (3), visualization It is structured to provide tools for the analysis and presentation of findings. Server (3) ensures that the model and data processing processes comply with the KVKK and other data protection laws. It is structured to ensure compliance with the laws. Server (3), It is configured to ensure data encryption and anonymization. 25 Server (3) regularly updates the model to market changes and new scams It is configured to ensure it is updated to meet the technical requirements. Server (3), feedback loop to monitor and improve model performance It is structured to ensure its establishment. 5 Industrial Application of the Invention Thanks to the system (1) that is the subject of the invention, the detection of suspicious transactions in accounting transactions, The risks of financial fraud and errors can be reduced, and sudden and unexpected financial losses can be minimized. Losses can be prevented, the company's financial security and reputation can be protected, manually By automating audit processes, personnel productivity is increased. Real-time detection and intervention increase transaction speed, identifying erroneous or suspicious items. Early detection of these issues reduces losses and legal process costs. Thanks to increased productivity, labor costs are decreasing, with AI-powered 10 Analytical reports provide valuable information for management and strategic planning, business Incentives are being created to encourage improvements in workflows and processes. Increasing the sense of financial security strengthens customer loyalty. Positive improvement in customer satisfaction thanks to increased service quality. This is provided by advanced analytical capabilities and market differentiation. 15 This ensures the company's leading position in the sector through fast and effective risk management. This contributes to their arrival, responding to the demands of regulatory bodies and oversight. compliance with standards, data protection laws such as KVKK (Personal Data Protection Law) and Compliance with international financial reporting standards is ensured continuously. This ensures the promotion of a culture of improvement and innovation. 20 Around these basic concepts, the subject of the invention is the “Smart Financial Audit System (1)” It is possible to develop a wide variety of related applications, and the invention described herein It cannot be limited to examples; it is essentially as stated in the requests.

Claims

6 REQUESTS 1. Using artificial intelligence algorithms, identify suspicious items in accounting data. and enables the detection of abnormal processes; - Data related to accounting transactions, artificial intelligence models and models 5 recording of training data and suspicious transaction reports used in training at least one database structured to ensure that it is kept under control (2) And - Collection and preparation of accounting data, identification of suspicious transactions Selecting and developing the model to be used, training the model 10 and verification, real-time suspicious transaction detection with the model. the implementation, reporting of detected actions, security measures to ensure that it is obtained and model performance is improved a system characterized by containing at least one configured server (3) (1). 15 2. Communicating with server (3) and exchanging data with server (3) characterized by the database structured to perform (2) A system like the one in claim 1 (1).

3. Containing a general ledger and a trial balance sheet. raw data obtained from sources, accounts receivable and accounts payable additional data collected from the records, selected artificial intelligence models training datasets to be used in training, real-time accounting Monitoring data related to the information revealed that 25 were identified as suspicious. Notifications created and reports prepared for transactions are on record. characterized by the database structured to ensure its retention (2) A system like the one in Request 1 or 2 (1). 7 4. To communicate with the database (2) and access the data in the database (2). the above characterized by the server (3) configured to access a system like any of the requests (1).

5. External servers containing general ledger and trial balance sheets. Server (3) configured to enable the collection of raw data a system like any of the above characterized claims (1).

6. Accessing accounts receivable and accounts payable records from external servers 10 characterized by the server (3) configured to collect additional information by providing a system like any of the above-mentioned requests (1).

7. Cleaning and normalizing the collected data and identifying missing data. 15 characterized by the server (3) configured to ensure its completion a system like any of the above-mentioned requests (1).

8. Enriching the dataset through feature engineering and necessary... with the server (3) configured to enable the extraction of features A system like any of the above-mentioned described requirements 20 (1).

9. Isolation Forest or One-Class for use in anomaly detection. To enable the selection of unsupervised algorithms in the form of SVMs. 25 of the above requests characterized by the configured server (3). a system like any other (1).

10. Random Forest, Gradient Boosting Machines, Deep Classification Server configured to enable the use of learning models. (3) like any of the above-mentioned claims characterized by system (1). 8 11. For time series analysis and ordinal models, use ARIMA, RNN, LSTM, or To enable the evaluation of algorithms in the form of GRUs. from the above requests characterized by the configured server (3) a system like any of them (1). 5 12. To ensure that the selected algorithms are trained on the training dataset. from the above requests characterized by the configured server (3) a system like any other (1).

13. Cross-validation and other validation techniques to enable evaluation of model performance using from the above requests characterized by the configured server (3) a system like any other (1).

14. To improve the model through hyperparameter tuning. from the above requests characterized by the configured server (3) a system like any other (1).

15. Integration of the developed model with existing accounting software 20 the above characterized by the server (3) configured to provide a system like any of the requests (1).

16. A job for automatically updating and retraining the model. 25 with server (3) configured to enable the establishment of the stream. a system like any of the above characterized claims (1).

17. Server (3), integration of the model into real-time data stream and continuous 30 characterized by the server (3) configured to enable monitoring. a system like any of the above requests (1). 9 18. Setting up alarm and notification systems for detecting suspicious transactions. to provide the above characterized by the configured server (3) a system like any of the requests (1).

19. To ensure that suspicious transactions are reported. from the above requests characterized by the configured server (3) a system like any other (1).

20. To enable the analysis and presentation of findings using visualization tools. from the above requests characterized by the configured server (3) a system like any other (1).

21. The model and data processing processes comply with the KVKK (Personal Data Protection Law) and other data protection laws. Characterized by the server (3) configured to ensure suitability 15 a system like any of the above-mentioned requests (1).

22. To ensure data encryption and anonymization. from the above requests characterized by the configured server (3) a system like any other (1). 20 23. The model is regularly updated to adapt to market changes and new scams. To ensure that the configured server (3) is updated against the techniques a system like any of the above characterized claims (1). 25 24. The feedback loop for monitoring and improving model performance. characterized by the server (3) configured to enable its establishment a system like any of the above requests (1).

Images