Payment reconciliation system and payment reconciliation program
The deposit reconciliation system uses machine learning to group and match billing data, addressing inefficiencies in identifying candidate combinations, thereby enhancing the efficiency of reconciliation processes.
Patent Information
- Application Number
- JP2023104993
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-06-27
- Publication Date
- 2025-01-16
AI Technical Summary
Existing systems struggle to efficiently identify appropriate combinations of billing data for reconciliation with deposit data in credit transactions, especially when there are multiple candidate combinations, leading to increased processing load and difficulty in determining the correct matches.
A deposit reconciliation system that groups billing data using machine learning and performs matching and reconciliation processes to efficiently identify candidate combinations, utilizing a CPU, OS, DBMS, and middleware to automate and streamline the reconciliation process.
The system efficiently identifies and reconciles candidate billing data combinations, reducing processing load and improving efficiency in matching complex billing and deposit data scenarios.
Smart Images

Figure 2025005014000001_ABST
Abstract
Description
[Technical field]
[0001] The present invention relates to a technique for clearing accounts receivable (deposit clearing) in credit transactions, and in particular to a technique that is effective when applied to a deposit clearing system and a deposit clearing program that matches billing data with deposit data. [Background technology]
[0002] In continuous commercial transactions between companies, credit transactions are conducted, and the amount invoiced (accounts receivable) is compared with the amount actually received at a later date, and if there are no discrepancies, the payment reconciliation work is carried out, in which the payment data is used to cancel the corresponding invoice data. This process was generally carried out manually by staff in the past, but in simple cases where, for example, payment data and invoice data correspond one-to-one, the reconciliation does not impose a large burden.
[0003] On the other hand, for example, there are cases where multiple invoices are issued to one customer, and payments are made for these multiple invoices together, with the granularity of the payments varying, and in such cases matching the invoice data with the payment data can be difficult and the load can be high. Also, even for combinations that appear to match, there may be cases where the amounts slightly differ due to processing of transfer fees, etc., and confirmation is required, which can increase the load.
[0004] In response to this situation, for example, JP 2021-96574 A (Patent Document 1) describes a technology for speeding up and streamlining the payment reconciliation process, which includes a means for selecting a combination of invoice data corresponding to payment data as reconciliation candidate data, and matching the reconciliation candidate data with the invoice data that has been learned after the reconciliation process, and converting the reconciliation candidate data and the learned invoice data into n-dimensional vectors based on invoice classification information, and calculating the degree of match between the reconciliation candidate data and the learned invoice data based on the cosine similarity between the two vectors, for example. [Prior art documents] [Patent documents]
[0005] [Patent Document 1] Patent Publication No. 2021-96574 Summary of the Invention [Problem to be solved by the invention]
[0006] According to the conventional technology described in Patent Document 1, when there are multiple candidate combinations of billing data for certain deposit data, it can be said that it is possible to efficiently determine which candidate combination is suitable as a target for clearing with the deposit data. However, there is no mention of how to specify the candidate combination of billing data to be cleared with the deposit data (how the billing data should be combined) before performing such a determination and clearing process.
[0007] SUMMARY OF THE PRESENT EMBODIMENT An object of the present invention is to provide a deposit reconciliation system and a deposit reconciliation program that make it possible to identify candidate combinations of billing data that will enable efficient reconciliation processing between billing data and deposit data.
[0008] The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings. [Means for solving the problem]
[0009] Among the inventions disclosed in this application, a brief summary of representative inventions is as follows.
[0010] A representative embodiment of the present invention is a deposit reconciliation system which performs reconciliation processing between billing data and deposit data relating to credit transactions, and includes a data input unit which receives and records the billing data and the deposit data, a billing grouping unit which groups the billing data that is presumed to be deposited collectively for each customer and divides it into billing groups, and a matching processing unit which matches the billing groups with the deposit data and performs reconciliation for the billing data included in the billing group that matches the deposit data.
[0011] Furthermore, the deposit reconciliation system which is a representative embodiment of the present invention can also be applied to a program for causing a computer to operate as the deposit reconciliation system described above. Effect of the Invention
[0012] The effects obtained by the representative inventions disclosed in this application can be briefly explained as follows.
[0013] In other words, according to a representative embodiment of the present invention, it is possible to efficiently identify candidate combinations of billing data to be reconciled with deposit data from a large amount of billing data, and to efficiently perform the deposit reconciliation process. [Brief description of the drawings]
[0014] [Figure 1] 1 is a diagram showing an overview of a configuration example of a payment clearing system according to an embodiment of the present invention. [Diagram 2] 1 is a flowchart outlining an example of the flow of a deposit clearing process in one embodiment of the present invention. [Diagram 3] FIG. 13 is a diagram outlining a specific example of deposit data according to an embodiment of the present invention. [Figure 4] A diagram showing an overview of a specific example of cancelled data in one embodiment of the present invention. [Diagram 5]FIG. 1 is a diagram showing an overview of an example of grouping of billing data in one embodiment of the present invention. [Figure 6] FIG. 13 is a diagram showing an overview of another example of grouping of billing data in one embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0015] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. In all the drawings for explaining the embodiment, the same parts are generally given the same reference numerals, and the repeated explanations will be omitted. On the other hand, a part that has been described with a reference numeral in a certain drawing may be referred to with the same reference numeral in the explanation of another drawing, although it will not be shown again.
[0016] <System configuration> Fig. 1 is a diagram showing an overview of an example of the configuration of a deposit reconciliation system according to one embodiment of the present invention. The deposit reconciliation system 1 of this embodiment is an information processing system that performs deposit reconciliation processing in the accounting and bookkeeping work of a company or the like, and may be implemented, for example, by an information processing terminal such as a PC (Personal Computer) or a tablet terminal, or by a server device or a virtual server constructed on a cloud computing service. Although the figure shows a single logical information processing system, the system may be physically implemented with its functions distributed across multiple information processing terminals, servers, etc.
[0017] The deposit reconciliation system 1 realizes various functions related to the deposit reconciliation process, for example by using a CPU (Central Processing Unit) not shown to execute middleware such as an OS (Operating System), a DBMS (DataBase Management System), and a Web server program that are expanded into memory from a recording device such as an HDD (Hard Disk Drive) or SSD (Solid State Drive), and software that runs on top of them.
[0018] This payment reconciliation system 1 has various units, such as a data input unit 20, a billing grouping unit 30, a name identification processing unit 40, a matching processing unit 50, an unmatch processing unit 60, a reconciliation processing unit 70, and a learning processing unit 80, which are implemented as software. These units may be implemented as one application program as a whole, or each unit may be implemented as an independent program and linked together. In addition, the payment reconciliation system 1 has various data stores, such as billing data 11, payment data 12, billing group data 13, reconciled data 14, and unreconciled data 15, which are implemented as databases, file tables, etc.
[0019] The data input unit 20 has a function of accepting input of billing data and deposit data to be the subject of the deposit clearing process, and storing them as billing data 11 and deposit data 12, respectively. The method of data input is not particularly limited, and may be, for example, a method of copying from a medium on which the data is recorded, or a method of uploading via a network (not shown). A method of directly manually inputting the data each time an invoice is issued or a deposit is made, or a method of scanning an invoice or the like to read the data, may also be used. A method of receiving generated data in cooperation with an accounting system, program, etc. (not shown) may also be used.
[0020] The invoice grouping unit 30 has a function of grouping invoice data that could not be cleared by matching by the name matching processing unit 40 described later, by machine learning processing using a learning model 81, and storing the grouped data as invoice group data 13. There are no particular limitations on the machine learning engine or algorithm, and generally available ones can be used as appropriate. The configuration may be such that grouping is performed manually by the user, without relying on machine learning processing. The details of the grouping processing will be described later.
[0021] The name matching processor 40 has a function of matching the billing data and the deposit data by customer, and performing a simple match between the billing data and the deposit data for each matched customer. For billing data and deposit data whose correspondence has been confirmed by matching, the reconciliation processor 70 (described later) performs a process related to the reconciliation of the billing data with the deposit data. Note that simple matching here corresponds to cases where there are no other candidates and the combination can be easily and uniquely identified, such as cases where the billing data and the deposit data correspond one-to-one, or cases where all the billing data and all the deposit data correspond together even if there are multiple data for either one or both.
[0022] The matching processing unit 50 has a function of matching the remaining unsettled billing data that could not be cleared by the matching in the name identification processing unit 40 with the unsettled deposit data for each group of billing data grouped by the above-mentioned billing grouping unit 30 and output to the billing group data 13. As for the billing data matched by matching (each billing data in a group) and the deposit data, the clearing processing unit 70 described later performs processing related to clearing, similar to the name identification processing unit 40.
[0023] In this case, for example, the billing group and the payment data are first simply matched 1:1, that is, the billing data in the billing group is aggregated and matched with the payment data, and the corresponding billing data in the billing group is cancelled using the matched payment data. After that, the remaining billing groups and payment data that cannot be cleared are matched in an M:1 combination. The remaining billing data in the billing group that cannot be cleared may be regrouped by the billing grouping unit 30 or manually, and the above-mentioned matching may be performed again. The process of matching with the billing group will be described later.
[0024] The unmatching processor 60 has a function of further matching the billing data and the deposit data that remain after they cannot be cleared by the name identification processor 40 and the matching processor 50. For example, when there are multiple candidates for a combination of billing data and deposit data that can be cleared and it is not possible to uniquely identify them, the possibility of clearing each candidate is scored using a predetermined method (for example, a technique such as that described in Patent Document 1 of the related art), and the top candidate can be automatically identified, or a list of scores can be presented to the user to allow manual identification.
[0025] In addition, in cases where a combination that seems to match does not match the amounts by a small amount, it is possible to determine whether or not to match automatically using AI or rule-based processing, or to present the user with a list of unreconciled invoice data and payment data and a list of the total amount, etc., so that the user can make a manual decision. For data that cannot be reconciled even after these processes, the list can be presented to the user so that the user can manually match it.
[0026] The reconciliation processing unit 70 has a function of reconciling the billing data with the deposit data by recording information on combinations of billing data and deposit data matched by the name identification processing unit 40, the matching processing unit 50, and the unmatch processing unit 60 in the reconciled data 14, and recording information on the remaining data that does not match in the unreconciled data 15. The billing data and deposit data recorded in the unreconciled data 15 are referred to as needed in the processing by the matching processing unit 50 and the unmatch processing unit 60.
[0027] The learning processing unit 80 has a function of generating and updating, by (re)learning from training data, a learning model 81 used when grouping billing data by machine learning processing in the billing grouping unit 30. At this time, for example, correct answer data consisting of a combination of billing data and payment data manually matched by the user in the unmatch processing unit 60 can be used as training data.
[0028] <Processing flow> FIG. 2 is a flow chart outlining an example of the flow of a payment reconciliation process in an embodiment of the present invention. When the process starts, the name matching processing unit 40 first performs name matching by customer name for the data recorded in the billing data 11 and the payment data 12 (S01). After that, for each matched customer, the name matching processing unit 40 performs a simple match between the billing data and the payment data, and the matched data is matched by the reconciliation processing unit 70 (S02). Note that, as described above, the simple match here corresponds to a case where there is no other candidate and the combination can be easily and uniquely identified, such as a case where the billing data and the payment data correspond 1:1, or a case where all the billing data and all the payment data correspond even if there are multiple data for either one or both.
[0029] Thereafter, the billing grouping unit 30 groups the unallocated billing data that could not be cleared by the simple matching in step S02 through machine learning processing using the learning model 81 (S03). Then, for each group of generated billing data, the matching processing unit 50 checks the data against unallocated deposit data, and the clearing processing unit 70 clears the data that matches (S04). Note that in this embodiment, the billing data that could not be cleared by the simple matching in step S02 is the subject of grouping, but the billing data may be grouped from the beginning without the simple matching in step S02.
[0030] Thereafter, as necessary, the unmatch processing unit 60 extracts candidates for combinations of billing data and deposit data that can be cleared from the billing data and deposit data remaining that could not be cleared in step S04, and presents a list of the clearing possibilities for each candidate to the user using a predetermined method (for example, a method such as that described in Patent Document 1 of the related art) (S05). Then, the user manually identifies candidates to be matched, and the clearing processing unit 70 performs the clearing processing (S06). As described above, the clearing processing may be performed automatically using the candidate with the highest score calculated in step S05.
[0031] Thereafter, re-learning is performed by machine learning processing using the combination of the group of billing data and the deposit data (correct answer data) specified by the user in step S06 as training data, and the learning model 81 used in step S03 is updated (S07), and the processing ends. The processing in step S07 may be performed separately at a predetermined timing asynchronously with the series of deposit settlement processing from steps S01 to S06.
[0032] <Example of payment application> FIG. 3 is a diagram outlining a specific example of deposit data 12 in one embodiment of the present invention. Deposit data 12 includes, for example, information on the billing destination (the customer who made the deposit), the actual date of deposit, and the amount of deposit. It is assumed that the deposit data does not include information such as a serial number that identifies the corresponding billing data. FIG. 4 is a diagram outlining a specific example of cleared data 14 in one embodiment of the present invention. Here, an example of billing data that has been cleared using the deposit data in the example of FIG. 3 is shown. The billing data includes, for example, information on the billing destination customer, the scheduled date of deposit and the amount of the bill, and the billing name indicating the contents of the bill.
[0033] For example, the top two rows of deposit data in the example in Figure 3 (both have a billing destination of "Customer A", a deposit date of 2023 / 05 / 01, and a total of 700,000 yen for the two deposits) can be used to clear the top three rows of invoice data in the example in Figure 4 (both have a billing destination of "Customer A", a scheduled payment date of 2023 / 05 / 01, and a total of 700,000 yen for the three invoices).This case is an example of an M:N match between deposit data and invoice data.
[0034] In addition, in the data for the invoice recipient "Customer B," the deposit date in the deposit data in Figure 3 differs from the expected deposit date in the invoice data in Figure 4 (the deposit date in the deposit data in Figure 3 is later), which indicates that there are customers who make deposits later than the expected deposit date assumed (or specified) at the time of invoice.
[0035] <Grouping of billing data> The following describes grouping of billing data by the billing grouping unit 30. In this embodiment, for billing data that does not match the deposit data 1:1, multiple billing data are grouped together and then matched with the deposit data. In other words, for each customer, the billing data is divided into groups that are likely to be deposit units by inferring the deposit data (how the deposits will be combined) from the contents and characteristics of the multiple billing data. This makes it possible to efficiently match even in complex cases where the billing data and deposit data match M:N (a group of billing data and a 1:N match of deposit data).
[0036] FIG. 5 is a diagram showing an overview of an example of grouping of billing data in one embodiment of the present invention. The upper diagram shows an example of billing data remaining in the unreconciled data 15. As described above, the payment data for each customer is estimated and divided into groups that are likely to be payment units. In this embodiment, as shown in the middle to lower diagrams, each billing data is encoded into an embedding vector, and the similarity between these vectors is evaluated and clustered by machine learning processing using a learning model 81. Then, each vector is inversely transformed for each cluster to return it to billing data, thereby obtaining the billing data contained in each group. Note that publicly known techniques, tools, etc. can be appropriately used for the means of embedding vectors and the means of clustering the obtained vectors by machine learning.
[0037] The embedding vector obtained from the billing data is a dimensional vector, and although each element and its value itself does not have any particular meaning, it is effective because it has high affinity with machine learning processing. As a result of grouping by machine learning, in the example of Figure 5, the billing data is divided into "Group 1" and "Group 2" as shown in the lower part, and by looking at the result, it can be inferred that the grouping was based on the billing date (2023 / 05 / 10 or 2023 / 05 / 11). In other words, since "Customer A" has had a strong tendency to receive payments in bulk on a billing date basis, it can be inferred that there is a high possibility that payments will be made in bulk on a billing date basis.
[0038] How the invoice data is grouped is up to the learning model 81, and unlike the above, for example, it is expected that the learning model 81 will be likely to group by product for a customer who has a strong tendency to make lump sum payments by product. On the other hand, if the units of lump sum payments are random with little regularity, it may not be possible to infer a trend even when looking at the grouping results after the fact because the machine learning process is a black box. Even in such cases, if the invoice data is grouped and compiled, it is likely that the matching process will be more efficient.
[0039] FIG. 6 is a diagram outlining another example of grouping of billing data in an embodiment of the present invention. The upper diagram shows an example of billing group data 13 related to unsettled billing data. Here, it is shown that each billing data is divided into "group 1" and "group 2" for each customer as a billing group. Looking at the result of this division, it can be inferred that "customer A" tends to make payments separately depending on whether the billing name is "maintenance fee" or something else, so that grouping was done by the scheduled payment date and the billing name ("maintenance fee" or something else). In addition, it can be inferred that "customer B" tends to make payments collectively one month after the billing, so that grouping was done by the scheduled payment date (one month later).
[0040] In other words, it can be seen that this grouping led to the assumption that there was a high possibility that payments would be made in units that were calculated by aggregating the billing data for each group. The bottom diagram in Figure 6 shows an example of payment data that is inferred from the grouping of billing data in the top diagram. If the actual payment data matches this, the payment data can be used to cancel the billing data in the corresponding billing group.
[0041] On the other hand, if the actual deposit data does not match (for example, if the deposit from "Customer B" on 2023 / 06 / 01 was ¥250,000 instead of ¥300,000), the user may manually make minor adjustments, such as by replacing the grouping of the billing data with other billing data, or the unsettled billing data may be regrouped by the billing grouping unit 30 to perform a new match.
[0042] As described above, according to the deposit reconciliation system 1 which is one embodiment of the present invention, by grouping and consolidating multiple invoice data based on inferences about the deposit data, it is possible to efficiently identify candidate combinations of invoice data that are subject to reconciliation with the deposit data from a large amount of invoice data, and efficiently perform reconciliation even in complex cases where invoice data and deposit data match M:N (a group of invoice data and a 1:N match).
[0043] The invention made by the present inventor has been specifically described above based on the embodiment, but it goes without saying that the present invention is not limited to the above embodiment and can be modified in various ways without departing from the gist of the invention. Furthermore, the above embodiment has been described in detail to explain the present invention in an easy-to-understand manner, and the present invention is not necessarily limited to having all of the configurations described. Furthermore, it is possible to add, delete, or replace part of the configuration of the above embodiment with other configurations.
[0044] In addition, the above-mentioned configurations, functions, processing units, processing means, etc. may be realized in part or in whole by hardware, for example, by designing them as integrated circuits. In addition, the above-mentioned configurations, functions, etc. may be realized in software by a processor interpreting and executing a program that realizes each function. Information such as the program, table, file, etc. that realizes each function can be stored in a recording device such as a memory, a hard disk, or an SSD, or in a recording medium such as an IC card, an SD card, or a DVD.
[0045] In addition, in each of the above figures, the control lines and information lines are shown as those considered necessary for explanation, and do not necessarily show all the control lines and information lines in the actual implementation. In reality, it can be considered that almost all the components are connected to each other. [Industrial Applicability]
[0046] The present invention can be used in a payment reconciliation system and a payment reconciliation program that matches billing data with payment data. [Explanation of symbols]
[0047] 1...Deposit reconciliation system, 11...Billing data, 12...Deposit data, 13...Billing group data, 14...Cleared data, 15...Uncleared data, 20... data input unit, 30... invoice grouping unit, 40... name identification processing unit, 50... matching processing unit, 60... unmatch processing unit, 70... settlement processing unit, 80... learning processing unit, 81…Learning Model
Claims
1. A payment reconciliation system that performs reconciliation processing between billing data and payment data related to credit transactions, a data input unit for receiving and recording the billing data and the deposit data; a billing grouping unit for grouping the billing data that is expected to be paid collectively for each of the customers into billing groups; A payment reconciliation system having a matching processing unit that compares the billing group with the payment data and performs reconciliation for the billing data included in the billing group that matches the payment data.
2. 2. The payment settlement system according to claim 1, The billing grouping unit is A payment reconciliation system that groups the billing data through machine learning processing using a learning model.
3. 3. The payment clearing system according to claim 2, The billing grouping unit is A payment reconciliation system that converts the billing data into embedded vectors and groups the data based on an evaluation of the similarity between the embedded vectors.
4. 3. The payment clearing system according to claim 2, Furthermore, an unmatching processing unit that accepts a combination of the billing data and the deposit data that is manually determined to match and performs allocation for the billing data and the deposit data that are not matched by the matching processing unit; A deposit reconciliation system comprising: a learning processing unit that re-learns the learning model based on the combination of the invoice data and the deposit data accepted by the unmatch processing unit.
5. A deposit reconciliation program that operates a computer as an information processing system that performs a reconciliation process between billing data and deposit data related to credit transactions, The computer, a first step of receiving and recording the invoice data and the deposit data; A second step of grouping the billing data that is expected to be paid collectively for each customer into billing groups; A third step of matching the billing group with the payment data and performing application for the billing data included in the billing group that matches the payment data.
Citation Information
Patent Citations
Deposit eliminating device, deposit eliminating method, and deposit eliminating program
JP2021096574A
