TR202100897A2 - Payroll control method and system - Google Patents

Abstract

The invention relates to a machine learning-based payroll control method and system that provides automatic detection of errors in source and input data that occur during the issuance of documents such as payroll showing the wages, taxes and deductions paid by the employer to the worker.

Description

DESCRIPTION Payroll control method and system Technical Area The invention shows the payroll, taxes and deductions paid by employers to their employees. automatically corrects errors in source and input data that occur during editing. machine learning-based payroll control method and system It is related to.

State of the Art The payroll is paid monthly for the work done by the employer for the employee. It is a document issued to show the fee paid, with taxes and deductions. Employer and all accounts between the employee and the employee are shown in detail. Burgundy, employer's It serves as proof that he/she has paid the wages he/she has to pay and he/she is responsible. Transactions must be arranged without errors by the department.

In the current technique, the work, personnel and payroll information provided by the employer as input and data such as output format are manually transferred to the personnel management platform by the operation. is entered as. The accuracy of all data is checked manually and the error When detected, it is manually corrected and the final output is as payroll and report. is offered.

The existing systems and algorithms in the current art can cause conventional and Although it can perform operational checks, it is accepted that the input data is correct. is working. Calculation that may occur with human hands or incorrect data entries unable to detect errors. Existing algorithms are only classical and rule set. Errors in source and input data can be made because it can make based checks. fails to reveal it. A patent application titled "system" has been found. Mentioned mobile reporting The feature of the system is that it is not browser-based and works on all platforms. mobile reporting system, accounting, human resources, payroll, stock, production and customer relations (CRM) by connecting to the database of package programs used in units. is reading data. Mobile reporting system by connecting to existing databases displays the required data cumulatively or in detail. However, in the application, the reporting system in question, which occurred during the issuance of the payroll. It is not mentioned that it provides automatic detection of errors.

As a result, due to the above-mentioned disadvantages and current solutions It is necessary to make an improvement in the relevant technical field due to the inadequacy of the is locked.

Purpose of the Invention The invention was created by being inspired by the current situations, and the above-mentioned negativities aims to solve.

The main purpose of the invention is to show the wages, taxes and deductions paid by the employer to his employee. source and input data generated during the editing of documents such as payroll. It is to ensure that errors are automatically detected through machine learning.

Another object of the invention is to increase work efficiency during payroll issuance.

The structural and characteristic features and all advantages of the invention are given in the following figures and It is clearer thanks to the detailed explanation written by making references to these figures. will be understood as such, and therefore the assessment will also include these figures and should be taken into account.

Figures to Help Understand the Invention Figure 1 is a representative flow diagram of the method of the invention.

Figure 2 is the block diagram view of the inventive system.

Description of Part References 1. Personnel management platform 2. Processor 3. Reporting unit Detailed Description of the Invention In this detailed description, payroll control based on machine learning, which is the subject of the invention, is presented. method and preferred embodiments of the system only explained for understanding.

The invention is like a payroll that shows the wages, taxes and deductions paid by the employer to the employee. errors in source and input data that occur during the editing of documents. machine learning-based payroll control system that provides automatic detection and It is a payroll control method performed using the system.

The system subject to the invention; . The work, personnel and work provided by the employer integrated by the operation A personnel management system where data such as payroll information and output format are entered. platform (1), o Required and personnel to perform payroll calculations machine that checks the accuracy of all data entered into the management platform a processor (2) on which learning algorithms are run, by clustering the payrolls of that processor (2) that exhibit similarity, within the model, taking into account the inter-period cluster changes, possible errors are detected. detects the error, either by editing the error or by the processor (1) error detection. If not, the reporting that creates the final document by continuing with the existing data. contains.

The method subject to the invention; 0 such as work, personnel and payroll information and output format provided by the employer data is integrated into the personnel management platform (1) by the operation. providing data entry for the current period; - in the information that is related to each other, logical control rules are applied and erroneous information Data entered into the minute platform to prevent regulation of traditional controls over it; o data so that the data evaluated in machine learning can be clustered digitizing and scaling their inputs; 0 method - distance - to determine which approach would be most appropriate identification and selection of cluster candidates; 0, on a yearly basis, of the clusters appearing in the approaches studied on a monthly basis. In order to be evaluated, it is necessary to keep similar elements together on a yearly basis. . creating an educated data set by determining the approach model; o by the processor (2) on which machine learning algorithms are run 0 that the accuracy of the data for the current period is beyond the traditional controls. control by operation, If the predicted error is detected as a result of 0 control, the error by correcting it or, if it is not detected, continuing with the existing data. report generation, 0 rechecking the generated preliminary report by the processor (2) and In case of detecting unforeseen errors as a result of the control If the error is corrected or not detected, continue with the existing data. preparation of the final report by the reporting unit (3) In a preferred embodiment of the system, which is the subject of the invention, the processor (2) also It also performs automatic comparison and trend analysis of the generated reports.

The processor (1), which is the subject of the invention, is required to perform payroll calculations. Check the accuracy of all data entered into the personnel management platform. is doing. These data, such as selected personnel, contract, position, wage, family information and There may be selected payment and deduction items. Here is an example working It is ensured that an incorrect entry that is set as "does not receive" is not allowed. in these information groups. differences result in differences in payroll. Personnel on payrolls There may be differences between the personnel and a personnel from period to period. Processor (1) possible errors by clustering similar payrolls, making a comparison model, it is determined by considering the cluster changes between periods.

Possible faults detected automatically by authorized personnel. examined, possible incorrect information pointed to is either corrected or an expected difference indicates that it is.

Digitizing and scaling the data makes all inputs [ 0 ; expression in the range of 1 ] is provided so that all payrolls of an institution are on a standardized basis. can be clustered. Example of counting: an employee's “Blue Collar - M” “White Collar - B” or “Intern - 8”: values M => 0.5 - trainee => 0 - B => 1 was given. Example of scaling: Monthly calculation of any payment or deduction item based on the minimum and maximum scaling - where min=0, max=1.

Method - distance - cluster to determine which approach would be most appropriate Candidates are identified and selected. Developing three criteria together (4 Methods - 5 Distance - 9 Cluster size) combinations are run and the outputs are examined (Silhouette Charts and Silhouette scores). Outputs produced by alternative approaches evaluated, and the selection of the Standard approach model, which will be specific to each institution, going. Alternative approaches for a minimum one-year information of the institution under consideration is run. In the determination of the strongest candidates, Silhouette graphs between months stability is sought (similarity) and Silhouette Scores close to 15 are evaluated first.

Another decision support application produced by the development is the selected Method and Distance. According to the Cluster number alternatives in the criteria, produced with fixed cluster names Anomaly report. By examining this output, it can be seen that produced in the past period. Assuming that payrolls are flawless, clustering with the least anomalies is preferred.

Maximum Silhouette Score, Minimum Anomaly based on the approach model. determines.

The subject of the invention is summarized as follows in its simplest form: all payrolls is subject to tests and when it passes the traditional tests successfully, The payrolls are filtered by removing the personnel entering and leaving the job. If the remaining payrolls is sent to the processor (2), and until all anomalies are examined, it is either an anomaly or A list of possible anomalies is generated as "not an anomaly".

Thanks to the invention, your report! resource created during the issuance of the payroll document and automatic detection of errors in input data through machine learning work efficiency during payroll arrangement increasing.

Claims (1)

REQUESTS It is a system to automatically detect the errors in the source and input data that occur during the preparation of documents such as payroll showing the wages, taxes and deductions paid by the employer to the employee. a personnel management platform (1) where data such as work, personnel and payroll information and output format provided by the employer are entered integratedly by that operation, - a machine learning algorithms run, which is necessary to perform payroll calculations and which checks the accuracy of all data entered into the personnel management platform If the processor (2) detects possible errors by clustering payrolls that exhibit similarity, within a comparison model, considering the cluster changes between periods, the error is edited or if the processor (1) does not detect an error, it continues with the existing data and creates the final document. includes reporting. It is a method to automatically detect the errors in the source and input data that occur during the preparation of documents such as payroll showing the wages, taxes and deductions paid by the employer to the employee. - providing data entry for the current period by integrating with the personnel management platform (1) by the operation, data provided by the employer, such as employment, personnel and payroll information and output format; o arranging traditional checks on the data entered into the minute platform in order to prevent erroneous information entry by operating logical control rules on information that is related to each other; - digitizing and scaling data entries so that data evaluated in machine learning can be aggregated; 0 method - distance - identification and selection of cluster candidates to determine which approach would be most appropriate; Fixing of clusters appearing in approaches run on a monthly basis; - creating an educated data set by determining the approach model; o Checking the accuracy of the data of the current period by the processor (2), on which machine learning algorithms are run, by running traditional controls, if the predicted error is detected as a result of 0 control, the error is corrected or if not, a preliminary report is created by continuing with the existing data, (2) and if an unforeseen error is detected as a result of the control, the error is corrected or if it is not detected, continuing with the existing data and creating the final report by the reporting unit (3). It is a method according to the claim, and its feature is; It is the automatic comparison of the reports generated by the processor (2). It is a method according to claim 25, its feature is; It is the creation of trend analysis based on the reports created by the processor (2). It is a method according to Claim Z, its feature is; Digitizing and scaling data is the process step of converting all inputs to a format that will be expressed in the [0;1] range. It is a method according to claim 2, and its feature is; It is the fixation of cluster names by keeping similar elements together on a yearly basis so that the clusters that appear in the approaches studied on a monthly basis can be evaluated on a yearly basis.