Financial reimbursement process optimization method based on intelligent reimbursement cloud platform

Abstract

The invention belongs to the field of financial reimbursement, and discloses a financial reimbursement process optimization method based on an intelligent reimbursement cloud platform, and the method comprises the steps: enabling a system to recognize the to-be-reimbursed invoice information uploaded by a user, automatically recognizing the invoice information through the OCR technology, forming an invoice text, and carrying out the comparison verification of the to-be-reimbursed invoice information and the invoice text. Performing risk identification and evaluation on the reimbursement application, determining a risk level, and automatically triggering a corresponding approval process according to the risk level; after the approval is passed, identifying the generated financial voucher information by using an OCR technology, comparing the financial voucher information with the to-be-reimbursed information and the invoice information, and if an exception is found, triggering a manual auditing process; and performing statistical analysis on the reimbursement data by using a big data technology, and performing visual display on an analysis result. According to the method, the OCR technology and the big data analysis means are integrated, the reimbursement risk level is intelligently identified and evaluated, the reimbursement data analysis capability is enhanced, and the overall efficiency and accuracy of the reimbursement process are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of financial reimbursement, and specifically is a financial reimbursement process optimization method based on an intelligent reimbursement cloud platform. Background Art

[0002] With the rapid development of information technology and the continuous expansion of enterprise scale, financial management, especially the automation and intelligence of reimbursement processes, has become an important means to improve the operational efficiency of enterprises. The traditional reimbursement process usually involves a large amount of paper document processing and manual review, which is not only inefficient but also prone to errors and delays. In recent years, with the application of advanced technologies such as optical character recognition technology, artificial intelligence, and big data analysis, the reimbursement process has gradually developed in the direction of digitalization and automation. The application of these technologies makes the financial reimbursement process more efficient, accurate, and convenient.

[0003] Although the existing reimbursement process automation technology has made certain progress, it still has some limitations. First, most reimbursement systems currently still face challenges in processing large-scale data, especially when processing highly concurrent reimbursement requests, the system's response speed and stability need to be improved; second, traditional reimbursement systems often lack effective risk identification and assessment mechanisms, and are unable to intelligently determine the risk level of reimbursement applications, resulting in some low-risk reimbursement applications also needing to go through cumbersome multi-level approval processes, adding unnecessary workload. Existing reimbursement systems are relatively weak in data analysis, and often lack the function of in-depth analysis and visualization of reimbursement data, making it difficult to provide timely and effective decision-making support for corporate management. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a financial reimbursement process optimization method based on an intelligent reimbursement cloud platform. By integrating OCR technology and big data analysis means, it can intelligently identify and evaluate the reimbursement risk level, thereby significantly improving the overall efficiency and accuracy of the reimbursement process.

[0005] The financial reimbursement process optimization method based on the intelligent reimbursement cloud platform described in the present invention comprises the following steps: Step 1: The system identifies the invoice information uploaded by the user, automatically identifies the invoice information using OCR technology, forms an invoice text, and compares and verifies the invoice information and the invoice text; Step 2: Build a risk assessment model to identify and assess the risks of reimbursement applications, determine the risk level, and automatically trigger the corresponding approval process based on the risk level; Step 3: After approval, use OCR technology to identify the generated financial voucher information and compare it with the information to be reimbursed and the invoice information; if any abnormality is found, the manual review process is triggered; Step 4: Use big data technology to conduct statistical analysis on the reimbursement data and visualize the analysis results.

[0006] Furthermore, in step 1, the invoice information to be reimbursed includes name, department, date, reason for reimbursement and reimbursement amount, and the invoice information includes invoice code, invoice number, invoice date, purchaser information, seller information, product information, amount, tax rate and tax amount.

[0007] Furthermore, step 2 is specifically as follows: Extract key information from reimbursement applications as features, including reimbursement amount, reimbursement type, reimbursement frequency, and reimbursement history of the reimbursement person; All features are normalized; The risk assessment model is established and expressed as: , In the formula, A nonlinear penalty term for scoring historical records, where is the adjustment parameter used to adjust the shape and strength of the nonlinear penalty term of the historical record score of the reimbursement person; R is the risk score of the reimbursement application, A is the reimbursement amount, F is the reimbursement frequency, H is the historical record score of the reimbursement person, T is the risk coefficient of the reimbursement type, P is the policy change factor, and A, F, H, T, and P are all standardized values; Set two thresholds and ,in , to divide the risk into three levels: low, medium and high; when When the number is 0, it means the risk of the reimbursement application is low and it will pass the approval process quickly; when When the number is , it indicates that there is a risk and further review is required; when When the number is , it indicates a high risk and requires strict review or reimbursement denial; According to the different risk levels, the corresponding approval path is automatically selected. When the risk score R exceeds the set high risk threshold When an emergency occurs, an early warning notification is automatically sent to the relevant departments.

[0008] Furthermore, in step 3, the financial voucher information includes the name of the person making the reimbursement, the reimbursement amount, and the reimbursement date; The anomaly detection formula for comparing the OCR recognition results with the original reimbursement information and invoice information is as follows: , in, The reimbursement amount identified by OCR; The reimbursement amount in the original reimbursement information; is the abnormality rate, which indicates the percentage difference in the reimbursement amount; Set the reimbursement abnormality threshold S based on historical reimbursement data; if If the reimbursement exception threshold S is exceeded, it is considered an exception and needs to enter the manual review process.

[0009] Furthermore, step 4 is specifically as follows: obtaining all completed reimbursement document data from the expense reimbursement system, preprocessing the collected data, storing the cleaned data in a structured database, and performing statistics and analysis on the data; selecting the corresponding chart type according to different data analysis requirements to display the reimbursement information of each department.

[0010] The beneficial effects described in the present invention are as follows: the present invention automatically identifies and verifies invoice information by applying OCR technology, thereby greatly reducing the need for manual intervention, significantly improving the automation level of the financial reimbursement process, and ensuring the accuracy and consistency of information; introducing an intelligent risk identification and assessment mechanism, automatically triggering the corresponding approval process according to the risk level of different reimbursement applications, realizing the refinement of risk management and the differentiation of approval processes, improving the approval efficiency while ensuring the security of the process; the present invention uses big data technology to conduct in-depth analysis of reimbursement data, and displays the analysis results in a visual manner, providing timely and accurate decision-making basis for corporate management, and helping to achieve more scientific and reasonable financial management. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a flow chart of the financial reimbursement process optimization method based on the intelligent reimbursement cloud platform in Example 1; Figure 2 A schematic diagram of risk identification and assessment for the reimbursement application in Example 1; Figure 3 This is a comparison chart of the recognition accuracy of each batch of reimbursement bills in Example 2. DETAILED DESCRIPTION

[0012] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments in conjunction with the accompanying drawings.

[0013] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a financial reimbursement process optimization method based on an intelligent reimbursement cloud platform, comprising the following steps:

[0014] Step 1: The system identifies the invoice information uploaded by the user, automatically identifies the invoice information using OCR technology, forms an invoice text, and compares and verifies the invoice information and the invoice text; The invoice information to be reimbursed includes name, department, date, reason for reimbursement and amount of reimbursement, and the invoice information includes invoice code, invoice number, invoice date, purchaser information, seller information, commodity information, amount, tax rate and tax amount; The user uploads an invoice image or layout file, and the system uses OCR technology to recognize the invoice image and layout file and extract the invoice information, including the invoice code, invoice number, invoice date, purchaser information, seller information, product information, amount, tax rate and tax amount. The OCR recognition result is verified to ensure the accuracy of the recognition information. The extracted invoice information is compared with the information to be reimbursed submitted by the user to ensure that the invoice information is consistent with the reimbursement content; Use the invoice QR code or format file information in the OCR recognition result to compare with the tax database to verify the authenticity of the invoice. Compare the invoice information with the information to be reimbursed to ensure that the invoice information is consistent with the reimbursement content, for example: amount comparison: whether the invoice amount is consistent with the amount to be reimbursed; product comparison: whether the invoice product information is consistent with the content to be reimbursed; time comparison: whether the invoice issuance date is consistent with the date to be reimbursed; purchaser comparison: whether the invoice purchaser information is consistent with the user information; Users fill in the information to be reimbursed online, including name, department, date, reason for reimbursement and amount to be reimbursed. The system automatically identifies the user's account information and extracts the information to be reimbursed. Use data cleaning technology to remove invalid or duplicate information, such as null values, abnormal values, duplicate invoice information and duplicate information to be reimbursed. Standardize the cleaned data, for example, unify the date format to YYYY-MM-DD and the amount format to numbers.

[0015] Step 2: Identify and assess the risks of reimbursement applications. Determine the risk level by following the steps below: Extract key information from reimbursement applications as features, including reimbursement amount, reimbursement type, reimbursement frequency, and reimbursement history of the reimbursement person; All features are normalized; The risk assessment model is established and expressed as: , In the formula, A nonlinear penalty term for scoring historical records, where is an adjustment parameter used to adjust the shape and strength of the nonlinear penalty term of the historical record score of the reimbursement person; R is the risk score of the reimbursement application, and its value range is [0,1]; A is the standardized value of the reimbursement amount, which reflects the size of the reimbursement amount relative to other reimbursement applications; F is the standardized value of the reimbursement frequency, which reflects the reimbursement frequency of the reimbursement person in a certain period of time; H is the standardized value of the historical record score of the reimbursement person, which is calculated based on the credibility of historical reimbursement behavior; T is the standardized value of the risk coefficient of the reimbursement type, and different types of reimbursements correspond to different risk coefficients; P is the standardized value of the policy change factor, which reflects the impact of the latest policy changes on the reimbursement risk; A, F, H, T, and P are all standardized values.

[0016] Set two thresholds and ,in , to divide the risk into three levels: low, medium and high; when When the number is 0, it means the risk of the reimbursement application is low and it will pass the approval process quickly; when When the number is , it indicates that there is a risk and further review is required; when When the number is , it indicates a high risk and requires strict review or reimbursement denial; According to the different risk levels, the corresponding approval path is automatically selected. When the risk score R exceeds the set high risk threshold When an emergency occurs, an early warning notification is automatically sent to the relevant departments.

[0017] Step 3: After approval, use OCR technology to identify the generated financial voucher information and compare it with the information to be reimbursed and the invoice information. If any abnormality is found, the manual review process is triggered; Financial voucher information includes the name of the person making the reimbursement, the reimbursement amount, and the reimbursement date; Use OCR technology to recognize the information on the generated financial vouchers. The recognition process includes image preprocessing, image segmentation, character recognition and recognition result processing; The anomaly detection formula for comparing the OCR recognition results with the original reimbursement information and invoice information is as follows: , in, The reimbursement amount identified by OCR; The reimbursement amount in the original reimbursement information; is the abnormality rate, which indicates the percentage difference in the reimbursement amount; By analyzing historical reimbursement data, the distribution of normal reimbursement amount differences is found. Usually, the average and standard deviation of the reimbursement amount differences in the historical data are calculated, and the reimbursement abnormality threshold S is set based on this; if If the reimbursement exceeds the abnormal threshold S, it is considered abnormal and needs to enter the manual review process; Manual review includes reviewing the original reimbursement information, invoice information and OCR recognition results. The reviewer will decide whether to approve the reimbursement or request additional materials based on the specific circumstances.

[0018] The value range of is [0,∞)%, which means the difference percentage of reimbursement amount. When it is close to 0%, it indicates that the OCR recognition result is consistent with the original reimbursement information and no additional review is required; When the reimbursement anomaly threshold S is reached or exceeded, it indicates a large discrepancy that may involve errors or fraud and requires further investigation.

[0019] Step 4: Use big data technology to perform statistical analysis on the reimbursement data and visualize the analysis results, including the following steps; Obtain all completed reimbursement document data from the expense reimbursement system; Preprocess the collected data, including removing duplicate records, filling missing values ​​and correcting erroneous data, and store the cleaned data in a structured database, such as using the relational database MySQL and the database MongoDB, and define fields such as the name of the person making the reimbursement, the reimbursement date, the reimbursement amount and the reimbursement category; calculate the total reimbursement amount, average reimbursement amount and number of reimbursements for each department, and use the big data analysis framework to perform complex statistical analysis tasks.

[0020] Suppose you need to calculate the total amount of reimbursements for a department within a specific time period, you can use the following formula: , Among them, B represents the total amount of reimbursement of a department in a specific time period. represents the amount of the i-th reimbursement, represents the date of the i-th reimbursement, n represents the total number of reimbursement records within the specified time period, is an indicator function, when The value is 1 if it falls within the specified time period, otherwise it is 0.

[0021] Use business intelligence tools such as Tableau and Power BI to create reports and dashboards. According to different data analysis needs, select appropriate chart types, such as bar charts, line charts, and pie charts, to display the total reimbursement amount, reimbursement trend, and reimbursement category distribution information of each department. In order to display the reimbursement trend, the moving average method can be used to smooth the data. The formula is as follows: , in, represents the moving average reimbursement amount at time point t, and the reimbursement amount at time point i is , Indicates the window width (i.e., number of days) used to calculate the moving average.

[0022] In summary, the financial reimbursement process optimization method based on the intelligent reimbursement cloud platform of the present invention not only significantly improves the efficiency of reimbursement processing, but also effectively reduces financial risks through automated risk assessment and anomaly detection mechanisms. In addition, the application of OCR technology greatly reduces manual input errors and improves the accuracy and completeness of reimbursement information. These advantages make the method of the present invention significantly more creative and novel than the traditional reimbursement process, bringing significant value to enterprises.

[0023] Example 2

[0024] Referring to Table 1, the second embodiment of the present invention is based on the first embodiment. In order to verify the effectiveness of the financial reimbursement process optimization method based on the intelligent reimbursement cloud platform proposed by the present invention, a three-month pilot project was conducted, and the financial reimbursement system of a large enterprise was selected as the research object. The enterprise processes an average of about 2,400 reimbursement documents per month, involving multiple departments and multiple types of reimbursements. Before implementing the present invention, the enterprise used a combination of traditional manual review and simple automation tools to process reimbursements.

[0025] First, the system is equipped with advanced OCR technology to automatically identify various invoice information uploaded by users, including electronic invoices and paper invoices. During the recognition process, the system extracts key information of each invoice, such as invoice code, invoice number and invoice date, and forms the invoice text. The invoice reimbursement information is compared and verified with the invoice text to ensure the consistency of the information.

[0026] The method of the present invention is used to identify and evaluate the risks of each reimbursement application to determine the risk level. Key information such as reimbursement amount, reimbursement type, reimbursement frequency, and historical records of the reimbursement person are extracted from the reimbursement application and standardized.

[0027] A risk assessment model is established to calculate the risk score of each reimbursement application. At the same time, two thresholds t1=0.2 and t2=0.6 are set to divide the risk levels into low, medium and high. According to the different risk levels, the most suitable approval path is automatically selected. For high-risk reimbursement applications, the system will automatically send early warning notifications to relevant departments.

[0028] For approved reimbursement applications, we use OCR technology to identify the generated financial voucher information again, extract the key information of each reimbursement application as features, standardize all features, apply the risk assessment model to calculate the risk score of each reimbursement application, and determine the risk level of the reimbursement application based on the risk score. The corresponding approval process is automatically triggered based on the risk level of the reimbursement application. If there is an anomaly between the OCR recognition result and the original reimbursement information, the manual review process is triggered.

[0029] Big data technology was used to statistically analyze all reimbursement data and visualize the analysis results. Reimbursement data from the past three months was collected, preprocessed, and stored in a structured database. Reimbursement information from different batches (about 400 copies per batch) was analyzed, and reimbursement trends and anomalies were displayed using charts.

[0030] Table 1 Experimental data record table

[0031] In this embodiment, by using the method of the present invention, the efficiency and accuracy of the financial reimbursement process are significantly improved. The specific analysis is as follows: OCR recognition accuracy: Figure 3 As shown in the figure, by adopting and optimizing OCR technology, the accuracy rate reaches over 97.5%, which is much higher than the accuracy rate of about 90% of traditional manual input. This greatly reduces the abnormal situations in the subsequent approval process and improves the smoothness of the overall process. Based on the risk assessment model of the present invention, reimbursement applications can be effectively divided into three risk levels: low, medium, and high. The results show that most reimbursement applications (numbers 001 to 003) have low risk scores (0.05 to 0.28), which indicates that the system can accurately screen out low-risk reimbursement applications and quickly pass the approval process. For high-risk reimbursement applications (numbers 005 and 006), the system can promptly identify and send early warning notifications so that relevant departments can conduct more detailed reviews; It can be seen from the data that the approval time increases slightly with the increase of the reimbursement amount, but overall, after adopting the method of the present invention, the approval time is shortened by about 30% compared with the traditional average approval time. This shows that through intelligent approval process, not only can the approval efficiency be improved, but also the labor cost can be reduced.

[0032] By using big data technology to conduct statistical analysis on reimbursement data and visualize the results, companies can more intuitively understand the reimbursement situation of each department, identify problems in a timely manner and take measures. For example, if the analysis finds that the reimbursement frequency of some departments is too high, there may be improper reimbursement behavior, and the company can adjust relevant policies or strengthen internal training accordingly.

[0033] The above description is only a preferred embodiment of the present invention and is not intended to be a further limitation of the present invention. All equivalent changes made using the contents of the present specification and drawings are within the protection scope of the present invention.

Claims

1. A financial reimbursement process optimization method based on an intelligent reimbursement cloud platform, characterized in that: The following steps are involved: Step 1: The system identifies the invoice information uploaded by the user, automatically identifies the invoice information using OCR technology, forms an invoice text, and compares and verifies the invoice information and the invoice text; Step 2: Build a risk assessment model to identify and assess the risks of reimbursement applications, determine the risk level, and automatically trigger the corresponding approval process based on the risk level; Specifically: Extract key information from reimbursement applications as features, including reimbursement amount, reimbursement type, reimbursement frequency, and reimbursement history of the reimbursement person; All features are normalized; The risk assessment model is established and expressed as: , In the formula, A nonlinear penalty term for scoring historical records, where is the adjustment parameter used to adjust the shape and strength of the nonlinear penalty term of the historical record score of the reimbursement person; R is the risk score of the reimbursement application, A is the reimbursement amount, F is the reimbursement frequency, H is the historical record score of the reimbursement person, T is the risk coefficient of the reimbursement type, P is the policy change factor, and A, F, H, T, and P are all standardized values; Set two thresholds and ,in , to divide the risk into three levels: low, medium and high; when When the number is 0, it means the risk of the reimbursement application is low and it will pass the approval process quickly; when When the number is , it indicates that there is a risk and further review is required; when When the number is , it indicates a high risk and requires strict review or reimbursement denial; According to the different risk levels, the corresponding approval path is automatically selected. When the risk score R exceeds the set high risk threshold Automatically send warning notifications to relevant departments; Step 3: After approval, use OCR technology to identify the generated financial voucher information and compare it with the information to be reimbursed and the invoice information; If an anomaly is found, the manual review process will be triggered; Step 4: Use big data technology to conduct statistical analysis on the reimbursement data and visualize the analysis results.

2. According to claim 1, a financial reimbursement process optimization method based on an intelligent reimbursement cloud platform is characterized in that: In step 1, the invoice information to be reimbursed includes name, department, date, reason for reimbursement and reimbursement amount, and the invoice information includes invoice code, invoice number, invoice date, purchaser information, seller information, product information, amount, tax rate and tax amount.

3. According to claim 1, a financial reimbursement process optimization method based on an intelligent reimbursement cloud platform is characterized in that: In step 3, the financial voucher information includes the name of the person making the reimbursement, the reimbursement amount, and the reimbursement date; The anomaly detection formula for comparing the OCR recognition results with the original reimbursement information and invoice information is as follows: , in, The reimbursement amount identified by OCR; The reimbursement amount in the original reimbursement information; is the abnormality rate, which indicates the percentage difference in the reimbursement amount; Set the reimbursement abnormality threshold S based on historical reimbursement data; if If the reimbursement exception threshold S is exceeded, it is considered an exception and needs to enter the manual review process.

4. The financial reimbursement process optimization method based on the intelligent reimbursement cloud platform according to claim 1 is characterized in that: Step 4 is as follows: Obtain all completed reimbursement document data from the expense reimbursement system; Preprocess the collected data, store the cleaned data in a structured database, and perform statistics and analysis on the data; Select chart types according to different data analysis needs to display the reimbursement information of each department.

Images