Anti-money laundering characterization parameter analysis system and method

Abstract

An anti-money laundering characterization parameter analysis system includes a processing unit and a storage unit electrically connected to the processing unit. The processing unit calculates an anomaly detection accuracy rate corresponding to each parameter based on the ratio between the number of anomalous cases corresponding to that parameter and the number of alert cases corresponding to that parameter. The processing unit also calculates an anomaly retention rate corresponding to each parameter based on the ratio between the number of anomalous cases corresponding to each parameter and a reference anomaly. The processing unit performs a screening procedure on the parameter data to select N most applicable parameter data points and generates and outputs a parameter screening result containing these N most applicable parameter data points.

Description

Technical Field

[0001] This invention relates to a parameter analysis system and method, and more particularly to a parameter analysis system and method for anti-money laundering characterization applicable to the field of anti-money laundering. Prior Technology

[0002] According to Article 9 of the "Regulations Governing the Prevention of Money Laundering by Financial Institutions" and Article 9 of the "Model Guidelines for the Prevention of Money Laundering by the Bankers Association," financial institutions should use risk-based approaches and information systems to detect suspected money laundering or terrorist financing transactions. Currently, financial institutions formulate relevant parameters to monitor transactions within the bank based on the anti-money laundering model established by the Financial Supervisory Commission. However, these parameters are based on past experience and may lead to an excessively high rate of normal transactions being mistakenly identified as money laundering or terrorist financing. Summary of the Invention

[0003] Therefore, one of the objectives of this invention is to provide an anti-money laundering characterization parameter analysis system applicable to the field of anti-money laundering.

[0004] This invention discloses an anti-money laundering characterization parameter analysis system, comprising a processing unit and a storage unit electrically connected to the processing unit. The storage unit stores multiple sets of parameter data, multiple alert case counts corresponding to the parameter data, and multiple abnormal case counts corresponding to the parameter data. For each set of parameter data and its corresponding alert case count, the alert case count indicates the number of alert transaction cases among the multiple sets of transaction case data, and each alert transaction case is one of the transaction case data that the computer determines has an abnormal risk based on the corresponding parameter data. For each set of parameter data and its corresponding abnormal case count, the abnormal case count indicates the number of abnormal transaction case data among the transaction case data, and each abnormal transaction case is one of the alert transaction case data corresponding to the parameter data that has been manually determined to be abnormal. The processing unit selects the largest abnormal case count as a reference abnormality value. The processing unit calculates an anomaly judgment accuracy rate corresponding to each parameter data based on the proportional relationship between the number of abnormal cases corresponding to each parameter data and the number of warning cases corresponding to each parameter data. The processing unit also calculates an abnormal case retention rate corresponding to each parameter data based on the proportional relationship between the number of abnormal cases corresponding to each parameter data and the reference anomaly quantity. The processing unit performs a filtering procedure on these parameter data to select N most applicable parameter data points and generates and outputs a parameter filtering result containing these N most applicable parameter data points. Where N is an integer greater than or equal to 1, the processing unit performs the screening procedure in the following ways: one of the two types of data, namely the anomaly judgment accuracy rate and the anomaly case retention rate, is used as a plurality of first cluster indicators, and the other type of data is used as a plurality of first screening indicators; the parameter data is divided into a plurality of groups according to the first cluster indicators, and one parameter data is selected from the parameter data of the group according to the first screening indicators corresponding to the parameter data of each group as a preliminary candidate parameter data; the N most applicable parameter data are determined from the preliminary candidate parameter data.

[0005] In some embodiments of the anti-money laundering characterization parameter analysis system of the present invention, the method by which the processing unit determines the N most applicable parameter data from the preliminary candidate parameter data includes: taking a plurality of first screening indicators corresponding to the preliminary candidate parameter data as a plurality of second grouping indicators, and taking a plurality of first grouping indicators corresponding to the preliminary candidate parameter data as a plurality of second screening indicators; dividing the preliminary candidate parameter data into a plurality of groups according to the second grouping indicators, and selecting one preliminary candidate parameter data from the preliminary candidate parameter data of each group as an advanced candidate parameter data according to the second screening indicators corresponding to the preliminary candidate parameter data of each group; and determining the N most applicable parameter data from the one or more advanced candidate parameter data.

[0006] In some embodiments of the anti-money laundering characterization parameter analysis system of the present invention, the processing unit performs the screening procedure by further including, before determining the N most applicable parameter data from the one or more advanced candidate parameter data: treating the first screening indicators as multiple third grouping indicators, and treating the first grouping indicators as multiple third screening indicators; dividing the parameter data into multiple groups according to the third grouping indicators; for the parameter data in each group, selecting one parameter data from the parameter data in that group as another preliminary candidate parameter data according to the third screening indicators corresponding to the parameter data in that group. The third screening indicators corresponding to the preliminary candidate parameter data are respectively used as multiple fourth grouping indicators, and the preliminary candidate parameter data are divided into multiple groups according to the fourth grouping indicators. For the preliminary candidate parameter data in each group, one preliminary candidate parameter data is selected from the preliminary candidate parameter data in that group as another advanced candidate parameter data according to the fourth screening indicators corresponding to the preliminary candidate parameter data in that group.

[0007] In some embodiments of the anti-money laundering characterization parameter analysis system of the present invention, the method by which the processing unit determines the N most applicable parameter data from one or more advanced candidate parameter data includes: determining whether the retention rate of the abnormal case corresponding to each advanced parameter data is greater than or equal to a retention rate threshold value, and when the determination result is yes, taking the advanced parameter data as one of the N most applicable parameter data.

[0008] In some embodiments of the anti-money laundering characterization parameter analysis system of the present invention, the storage unit also stores transaction case data. Each transaction case data includes a transaction amount and an attribute tag. Each parameter data includes multiple amount threshold values, each corresponding to a case attribute feature. The processing unit performs a verification process on each transaction case data based on each parameter data. Specifically, the processing unit performs the verification process by using the attribute tag of the transaction case data as the basis for determining the amount threshold value corresponding to the case attribute feature in the parameter data that matches the attribute tag as a matching amount threshold value, comparing the transaction amount of the transaction case data with the matching amount threshold value, and, if it is determined that the transaction amount is greater than or equal to the matching amount threshold value, designating the transaction case data as a warning transaction case data corresponding to the parameter data.

[0009] The processing unit counts the number of alert transaction cases corresponding to each parameter data to obtain the alert case quantity corresponding to the parameter data, and stores the alert case quantity in the storage unit. For each parameter data, the processing unit treats each alert transaction case with an abnormal annotation as an abnormal transaction case corresponding to the parameter data as an abnormal transaction case, counts the number of abnormal transaction cases corresponding to each parameter data to obtain the abnormal case quantity corresponding to the parameter data, and stores the abnormal case quantity in the storage unit.

[0010] Another objective of this invention is to provide a method for analyzing anti-money laundering characterization parameters in the aforementioned anti-money laundering characterization parameter analysis system.

[0011] This invention discloses a method for analyzing anti-money laundering characterization parameters, implemented by an anti-money laundering characterization parameter analysis system. This system stores multiple sets of parameter data, multiple alert case counts corresponding to these parameter data, and multiple abnormal case counts corresponding to these parameter data. For each set of parameter data and its corresponding alert case count, the alert case count indicates the number of alert transaction cases among the multiple sets of transaction case data, and each alert transaction case is one of those transaction cases identified by the computer as having an abnormal risk based on the corresponding parameter data. For each set of parameter data and its corresponding abnormal case count, the abnormal case count indicates the number of abnormal transaction cases among the multiple sets of transaction case data, and each abnormal transaction case is one of those transaction cases that matches the parameter data and is manually identified as abnormal. This anti-money laundering characterization parameter analysis method includes:

[0012] (A) The anti-money laundering characterization parameter analysis system selects the largest abnormal case from the number of such cases as a reference abnormality; (B) The anti-money laundering characterization parameter analysis system calculates an anomaly judgment accuracy rate corresponding to the parameter data based on the proportional relationship between the number of abnormal cases corresponding to each parameter data and the number of warning cases corresponding to the parameter data; (C) The anti-money laundering characterization parameter analysis system calculates an abnormal case retention rate corresponding to the parameter data based on the proportional relationship between the number of abnormal cases corresponding to each parameter data and the reference abnormality;

[0013] (D) The anti-money laundering characterization parameter analysis system performs a screening procedure on the parameter data to select N most applicable parameter data from the parameter data, and generates and outputs a parameter screening result containing the N most applicable parameter data, where N is an integer greater than or equal to 1. The anti-money laundering characterization parameter analysis system performs the screening procedure in the following ways: (D1) using one of the two types of data, namely the anomaly judgment accuracy rate and the anomaly case retention rate, as multiple first cluster indicators, and using the other type of data as multiple first screening indicators; (D2) dividing the parameter data into multiple groups according to the first cluster indicators, and selecting one parameter data from the parameter data in each group as a preliminary candidate parameter data according to the first screening indicators corresponding to the parameter data in each group; and (D3) determining the N most applicable parameter data from the preliminary candidate parameter data.

[0014] In some embodiments of the anti-money laundering characterization parameter analysis method of the present invention, in step (D3), the method by which the anti-money laundering characterization parameter analysis system determines the N most applicable parameter data from the preliminary candidate parameter data includes: (D31) taking a plurality of first screening indicators corresponding to the preliminary candidate parameter data as a plurality of second grouping indicators, and taking a plurality of first grouping indicators corresponding to the preliminary candidate parameter data as a plurality of second screening indicators; (D32) dividing the preliminary candidate parameter data into a plurality of groups according to the second grouping indicators, and selecting one preliminary candidate parameter data from the preliminary candidate parameter data of the group as an advanced candidate parameter data according to the second screening indicators corresponding to the preliminary candidate parameter data of each group; and (D33) determining the N most applicable parameter data from the one or more advanced candidate parameter data.

[0015] In some embodiments of the anti-money laundering characterization parameter analysis method of the present invention, in step (D3), the method by which the anti-money laundering characterization parameter analysis system performs the screening procedure further includes, before determining the N most applicable parameter data from one or more advanced candidate parameter data: (D34) treating the first screening indicators as multiple third grouping indicators, and treating the first grouping indicators as multiple third screening indicators; (D35) dividing the parameter data into multiple groups according to the third grouping indicators, and for the parameter data in each group, selecting one parameter data from the parameter data in that group as another preliminary candidate parameter data according to the third screening indicators corresponding to the parameter data in that group. (D36) The third screening indicators corresponding to the preliminary candidate parameter data are respectively used as multiple fourth grouping indicators, and the third grouping indicators corresponding to the preliminary candidate parameter data are respectively used as multiple fourth screening indicators; and (D37) The preliminary candidate parameter data are divided into multiple groups according to the fourth grouping indicators. For the preliminary candidate parameter data in each group, the anti-money laundering characterization parameter analysis system selects one preliminary candidate parameter data from the preliminary candidate parameter data in the group as another advanced candidate parameter data according to the fourth screening indicators corresponding to the preliminary candidate parameter data in the group.

[0016] In some embodiments of the anti-money laundering characterization parameter analysis method of the present invention, in step (D33), the method by which the anti-money laundering characterization parameter analysis system determines the N most applicable parameter data from one or more advanced candidate parameter data includes: determining whether the retention rate of the abnormal case corresponding to each advanced parameter data is greater than or equal to a retention rate threshold value, and when the determination result is yes, taking the advanced parameter data as one of the N most applicable parameter data.

[0017] In some embodiments of the anti-money laundering characterization parameter analysis method of the present invention, the anti-money laundering characterization parameter analysis system also stores such transaction case data. Each transaction case data includes a transaction amount and an attribute marker. Each parameter data includes multiple amount threshold values, each corresponding to a case attribute feature. The anti-money laundering characterization parameter analysis method further includes, prior to step (A): (E) The anti-money laundering characterization parameter analysis system performs a verification process on each transaction case data according to each parameter data, wherein the anti-money laundering characterization parameter analysis system performs the verification process in the following manner: (E1) Based on the attribute marker of the transaction case data, the amount threshold value corresponding to the case attribute feature in the parameter data that matches the attribute marker is used as a matching amount threshold value; (E2) The transaction amount of the transaction case data is compared with the matching amount threshold value; (E3) When it is determined that the transaction amount is greater than or equal to the matching amount threshold value, the transaction case data is used as a warning transaction case data corresponding to the parameter data;

[0018] (F) The anti-money laundering characterization parameter analysis system counts the number of alert transaction cases corresponding to each parameter data to obtain the number of alert cases corresponding to the parameter data, and stores the number of alert cases; and (G) For each parameter data, the anti-money laundering characterization parameter analysis system takes each alert transaction case with an abnormal annotation in the alert transaction case data corresponding to the parameter data as an abnormal transaction case data corresponding to the parameter data, counts the number of abnormal transaction case data corresponding to each parameter data to obtain the number of abnormal cases corresponding to the parameter data, and stores the number of abnormal cases.

[0019] The advantages of this invention are as follows: the anti-money laundering characterization parameter analysis system uses the processing unit to use one of the data, namely the accuracy rate of anomaly judgment and the retention rate of anomaly cases, as multiple first grouping indicators, and the other data as multiple first screening indicators, to select a parameter data from the parameter data in the group as the preliminary candidate parameter data, and then to determine the N most applicable parameter data from the preliminary candidate parameter data, so that the processing unit can improve the accuracy of finding money laundering or terrorist financing transaction data by using the N most applicable parameter data. Simple Explanation of the Diagram

[0020] Other features and effects of the present invention will be clearly presented in the embodiments with reference to the drawings.

[0021] Figure 1 is a block diagram illustrating an embodiment of the anti-money laundering characterization parameter analysis system of the present invention.

[0022] Figure 2 is a flowchart illustrating, by way of example, how an anti-money laundering characterization parameter analysis method is implemented in this embodiment.

[0023] Figure 3 is a flowchart illustrating, by way of example, how a check process is implemented in this embodiment.

[0024] Figure 4 is a flowchart illustrating, by way of example, how a screening procedure is implemented in this embodiment. Implementation

[0025] Before this invention is described in detail, it should be noted that, unless otherwise defined, the term "electrically connected" in this patent specification refers to the "coupled" relationship between computer hardware (e.g., electronic systems, devices, apparatuses, units, components), and broadly refers to "wired electrical connections" achieved by physically connecting multiple computer hardware components through conductor / semiconductor materials, and "radio connections" that achieve wireless data transmission using wireless communication technologies (e.g., but not limited to wireless networks, Bluetooth, and electromagnetic induction). On the other hand, unless otherwise defined, the term "electrical connection" in this patent specification also broadly refers to "direct electrical connections" achieved by directly coupling multiple computer hardware components to each other, and "indirect electrical connections" achieved by indirectly coupling multiple computer hardware components through other computer hardware components.

[0026] Before this invention is described in detail, it should be noted that the term "unit" in this patent specification refers to computer hardware, not software. For example, "processing unit" refers to computer hardware with data processing capabilities. Furthermore, the term "unit" in this patent specification can refer to a single piece of computer hardware with a specific function, or it can refer to a group of computer hardware with similar functions. For example, "processing unit" can refer to a single processor with data processing capabilities, but it can also refer to a collection of processors.

[0027] Referring to Figure 1, one embodiment of the anti-money laundering characterization parameter analysis system 100 of the present invention is a servo device, and the anti-money laundering characterization parameter analysis system 100 includes a processing unit 1 and a storage unit 2 electrically connected to the processing unit 1.

[0028] In this embodiment, the processing unit 1 is a processor implemented with integrated circuitry and capable of data processing and instruction transmission / reception. However, in different embodiments, the processing unit 1 may also be a circuit assembly including a processor and a circuit board, or a collection of multiple processors.

[0029] The storage unit 2 is a data storage device (e.g., a hard disk) used to store digital data. However, in different embodiments, the storage unit 2 may also be a collection of multiple storage devices of the same or different types.

[0030] The storage unit 2 stores multiple transaction case data 21 and multiple parameter data 22.

[0031] Each transaction case data 21 corresponds to a historical financial transaction (e.g., deposit, withdrawal, transfer, or remittance transaction) and includes a transaction amount and an attribute tag. At least one of these transaction case data 21 also includes an anomaly annotation. For example, the transaction amount is the amount deposited on that day. The attribute tag indicates, for example, one of "high-risk legal entity," "moderate-risk legal entity," "high-risk natural person," and "moderate-risk natural person," but is not limited to these categories. For each transaction case data 21, if the transaction case data 21 contains the anomaly annotation, it indicates that the historical financial transaction corresponding to the transaction case data 21 has been identified as a money laundering or terrorist financing transaction.

[0032] Each parameter data 22 includes multiple monetary threshold values, each corresponding to a case attribute feature. In this embodiment, each parameter data 22 includes four monetary threshold values. These four monetary threshold values ​​are respectively used as four threshold values ​​for the amount deposited on the same day, and each corresponds to one of the four case attribute features. For ease of illustration, two of these parameter data 22 are referred to here as a first parameter data and a second parameter data, respectively.

[0033] For example, the first parameter data includes four monetary thresholds of 1 million, 3 million, 500,000, and 1.5 million. The case attribute characteristic corresponding to the monetary threshold of "1 million" is "high-risk legal person," the case attribute characteristic corresponding to the monetary threshold of "3 million" is "generally risky legal person," the case attribute characteristic corresponding to the monetary threshold of "500,000" is "high-risk natural person," and the case attribute characteristic corresponding to the monetary threshold of "1.5 million" is "generally risky natural person." As another example, the second parameter data includes four monetary thresholds of 2 million, 4 million, 1 million, and 2.5 million. The case with a threshold of "2 million" is characterized as a "high-risk legal person", the case with a threshold of "4 million" is characterized as a "generally risky legal person", the case with a threshold of "1 million" is characterized as a "high-risk natural person", and the case with a threshold of "2.5 million" is characterized as a "generally risky natural person".

[0034] The first parameter data and the second parameter data are shown in Table 1 below. Table 1 The corresponding monetary threshold for "high-risk legal entities" The corresponding amount threshold for "general risk legal entities" The corresponding monetary threshold for "high-risk individuals" The corresponding monetary threshold for "general risk natural persons" First parameter data 1 million 3 million 500,000 1.5 million Second parameter data 2 million 4 million 1 million 2.5 million

[0035] It should be further noted that, in different embodiments, each parameter data 22 may include only one, two, or other different numbers of amount threshold values. Moreover, each amount threshold value may be, for example, a threshold value for withdrawal, transfer, or remittance amounts. Furthermore, the case attribute characteristics corresponding to each amount threshold value can be freely defined in conjunction with the attribute tags included in each transaction case data 21. Therefore, the specific actual form of such parameter data 22 is not limited to this embodiment.

[0036] Referring to Figures 1 and 2, the following exemplarily illustrates how the anti-money laundering characterization parameter analysis system 100 of this embodiment implements an anti-money laundering characterization parameter analysis method.

[0037] First, in step S01, the processing unit 1 performs a check on each transaction case data 21 based on each parameter data 22. This check is used to enable the processing unit 1 to use each parameter data 22 as a screening criterion to determine whether the transaction case data 21 has any abnormal risks (such as money laundering or terrorist financing risks).

[0038] Referring further to Figures 1 and 3, the following explains how the anti-money laundering characterization parameter analysis system 100 implements this verification process.

[0039] First, in step S11, the processing unit 1 uses the attribute tag of the transaction case data 21 as the matching amount threshold value for the case attribute feature in the parameter data 22 that matches the attribute tag. Continuing the example above, assuming the attribute tag of the transaction case data 21 indicates "general risk natural person," for the first parameter data, the processing unit 1 uses the amount threshold value of "1.5 million" as the matching amount threshold value. For the second parameter data, the processing unit 1 uses the amount threshold value of "2.5 million" as the matching amount threshold value. Next, the process proceeds to step S12.

[0040] In step S12, the processing unit 1 compares the transaction amount in the transaction case data 21 with the matching amount threshold to determine whether the transaction amount is greater than or equal to the matching amount threshold. Then, the process proceeds to step S13.

[0041] In step S13, when processing unit 1 determines that the transaction amount is greater than or equal to the matching amount threshold, it treats transaction case data 21 as a warning transaction case data corresponding to parameter data 22. Continuing the example above, assuming the transaction amount and attribute markers of transaction case data 21 are "2 million yuan" and "general risk individual" respectively, processing unit 1 will treat transaction case data 21 as one of the warning transaction case data corresponding to the first parameter data, but will not treat transaction case data 21 as a warning transaction case data corresponding to the second parameter data. In other words, if the first parameter data is used as the screening criterion, processing unit 1 will determine that transaction case data 21 has abnormal risk, but if the second parameter data is used as the screening criterion, processing unit 1 will determine that transaction case data 21 does not have abnormal risk.

[0042] The above is an example of how the anti-money laundering characterization parameter analysis system 100 in this embodiment implements the verification process.

[0043] After the check is performed in processing unit 1, the process proceeds to step S02.

[0044] In step S02, the processing unit 1 counts the number of alert transaction cases corresponding to each parameter data 22 to obtain a number of alert cases corresponding to each parameter data 22, and stores the number of alert cases in the storage unit 2. Each number of alert cases corresponds to the number of transaction cases 21 that the anti-money laundering characterization parameter analysis system 100 determines to have an abnormal risk based on the corresponding parameter data 22.

[0045] After the processing unit 1 stores the number of alert cases to the storage unit 2, the process proceeds to step S03.

[0046] In step S03, for each parameter data 22, the processing unit 1 treats each warning transaction case with the abnormal annotation in the warning transaction case data corresponding to the parameter data 22 as an abnormal transaction case data corresponding to the parameter data 22, counts the number of abnormal transaction case data corresponding to each parameter data 22 to obtain an abnormal case quantity corresponding to the parameter data 22, and stores the abnormal case quantity in the storage unit 2. Each abnormal case quantity indicates the number of transaction case data 21 that matches the corresponding parameter data 22 and is manually determined to be abnormal (e.g., belonging to money laundering or terrorist financing transactions).

[0047] After the processing unit 1 stores the number of abnormal cases to the storage unit 2, the process proceeds to step S04.

[0048] In step S04, the processing unit 1 selects the largest one from the abnormal case quantities as a reference abnormal quantity. Then, the process proceeds to step S05.

[0049] In step S05, the processing unit 1 calculates an anomaly judgment accuracy rate corresponding to parameter data 22 based on the proportional relationship between the number of abnormal cases and the number of warning cases corresponding to each parameter data 22. More specifically, the processing unit 1 divides the number of abnormal cases corresponding to parameter data 22 by the number of warning cases corresponding to parameter data 22 and takes the percentage as the anomaly judgment accuracy rate corresponding to parameter data 22. For example, the number of abnormal cases corresponding to the first parameter data is, for example, 136, and the number of warning cases corresponding to the first parameter data is, for example, 3354. The processing unit 1 divides 136 by 3354 and takes the percentage to calculate 4.05%, then "4.05%" is the anomaly judgment accuracy rate corresponding to the first parameter data.

[0050] After the processing unit 1 calculates the accuracy rate of such anomaly detection, the process proceeds to step S06.

[0051] In step S06, the processing unit 1 calculates an abnormal case retention rate corresponding to the parameter data 22 based on the proportional relationship between the number of abnormal cases corresponding to each parameter data 22 and the reference abnormal number. More specifically, the processing unit 1 divides the number of abnormal cases corresponding to the parameter data 22 by the reference abnormal number and takes the percentage as the abnormal case retention rate corresponding to the parameter data 22. Continuing the example above, the reference abnormal number is, for example, 136 cases. The processing unit 1 divides 136 cases by 136 cases and takes the percentage to calculate 100%, then "100%" is the abnormal case retention rate corresponding to the first parameter data.

[0052] After the processing unit 1 calculates the retention rate of such abnormal cases, the process proceeds to step S07.

[0053] In step S07, the processing unit 1 performs a filtering procedure on the parameter data 22 to select N most applicable parameter data from the parameter data 22, and generates and outputs a parameter filtering result containing the N most applicable parameter data. Here, N is an integer greater than or equal to 1. The N most applicable parameter data can be used by the processing unit 1 to filter new transaction case data (i.e., perform the verification process), thereby improving the accuracy of the processing unit 1 in identifying suspected money laundering or terrorist financing cases from the transaction case data. It should be noted that the processing unit 1 can output the parameter filtering result by controlling a display screen (not shown) to display the parameter filtering result, transmitting the parameter filtering result to a user terminal device (not shown), or providing the parameter filtering result to the storage unit 2 for storage.

[0054] Referring further to Figures 1 and 4, the following exemplarily illustrates how the anti-money laundering characterization parameter analysis system 100 of this embodiment implements the screening procedure. Steps S21 to S24 are considered a first type of screening process in this embodiment, and steps S25 to S28 are considered a second type of screening process in this embodiment.

[0055] First, in step S21, the processing unit 1 uses one of the data—the anomaly detection accuracy rate and the anomaly case retention rate—as multiple first clustering indicators, and uses the other data as multiple first screening indicators. More specifically, in this embodiment, the processing unit 1 uses the anomaly case retention rate as the first clustering indicator and the anomaly detection accuracy rate as the first screening indicator.

[0056] In step S22, the processing unit 1 divides the parameter data 22 into multiple groups of parameter data 22 according to the first clustering index, and selects one parameter data 22 from the parameter data 22 of each group as a preliminary candidate parameter data corresponding to the first type of screening process according to the first screening index corresponding to the parameter data 22 of each group. More specifically, in this embodiment, the processing unit 1 groups multiple parameter data 22 whose abnormal case retention rates (i.e., the first clustering index) are equal to each other. In other words, the abnormal case retention rates (i.e., the first clustering index) corresponding to the parameter data 22 of the same group are equal to each other. For each group of parameter data 22, the processing unit 1 selects the parameter data 22 with the highest corresponding abnormal judgment accuracy from the parameter data 22 of the group as the preliminary candidate parameter data corresponding to the first type of screening process.

[0057] Table 2 below exemplarily shows the accuracy rates for six anomaly detections and the retention rates for six anomaly cases corresponding to six of the six parameter data points 22. These six parameter data points are, for example, the first parameter data, the second parameter data, a third parameter data, a fourth parameter data, a fifth parameter data, and a sixth parameter data. Table 2 Parameter data number Anomaly detection accuracy Abnormal case retention rate First parameter data 4.05% 100% Second parameter data 4.00% 100% Third parameter data 3.95% 100% Fourth parameter data 4.05% 98.53% Fifth parameter data 4.00% 98.53% Sixth parameter data 3.95% 98.53%

[0058] Taking Table 2 as an example, the abnormal case retention rate (i.e., the first clustering index) corresponding to the first to third parameter data is 100%, and the abnormal case retention rate corresponding to the fourth to sixth parameter data is 98.53%. Therefore, the processing unit 1 will group the first to third parameter data into the same group 22 (e.g., the parameter data 22 in the first group), and the fourth to sixth parameter data into another group 22 (e.g., the parameter data 22 in the second group). Further, for the parameter data 22 in the first group, the processing unit 1 will select the one with the highest corresponding abnormality judgment accuracy (i.e., the first screening index) (i.e., the first parameter data) as one of the preliminary candidate parameter data corresponding to the first type of screening process. On the other hand, for the parameter data 22 in the second group, the processing unit 1 will select the one with the highest corresponding abnormality judgment accuracy (i.e., the fourth parameter data) as another preliminary candidate parameter data corresponding to the first type of screening process.

[0059] Furthermore, in another embodiment, in step S21, the processing unit 1 uses the anomaly detection accuracy rate as the first grouping index and the anomaly case retention rate as the first screening index. Therefore, in step S22 of this other embodiment, the processing unit 1 groups multiple parameter data 22 with equal anomaly detection accuracy rates into the same group, and then uses the parameter data 22 in the group with the highest anomaly case retention rate as the preliminary candidate parameter data corresponding to the first type of screening process.

[0060] After the processing unit 1 uses the first clustering index and the first screening index to select the preliminary candidate parameter data corresponding to the first type of screening process, the process proceeds to step S23.

[0061] In step S23, the processing unit 1 uses multiple first screening indicators that correspond to the preliminary candidate parameter data corresponding to the first type of screening process as multiple second clustering indicators, and uses multiple first clustering indicators that correspond to the preliminary candidate parameter data as multiple second screening indicators. More specifically, in this embodiment, the processing unit 1 uses the anomaly judgment accuracy rate corresponding to the preliminary candidate parameter data corresponding to the first type of screening process as the second clustering indicators, and uses the abnormal case retention rate corresponding to the preliminary candidate parameter data corresponding to the first type of screening process as the second screening indicators.

[0062] In step S24, the processing unit 1 divides the preliminary candidate parameter data corresponding to the first type of screening into multiple groups corresponding to the first type of screening based on the second clustering index. Furthermore, based on the second screening index corresponding to each group of preliminary candidate parameter data corresponding to the first type of screening, the processing unit 1 selects one preliminary candidate parameter data from that group as an advanced candidate parameter data corresponding to the first type of screening. More specifically, in this embodiment, the processing unit 1 groups multiple preliminary parameter data whose anomaly detection accuracy (i.e., the second clustering index) is equal among all the preliminary parameter data corresponding to the first type of screening. In other words, the anomaly detection accuracy (i.e., the second clustering index) corresponding to the preliminary parameter data in the same group are equal. For each group of parameter data 22, the processing unit 1 selects the one with the largest corresponding second screening index as the advanced candidate parameter data corresponding to the first type of screening.

[0063] Following the example in Table 2 above, the first and fourth parameter data points have both been used as preliminary candidate parameter data points corresponding to the first type of screening process in step S22, and their corresponding anomaly judgment accuracy rates (i.e., the second grouping index) are both 4.05%. Therefore, the processing unit 1 will classify the first parameter data point and the fourth parameter data point into the same group of preliminary candidate parameter data points corresponding to the first type of screening process. Further, for the preliminary candidate parameter data points in this group, the processing unit 1 will select the one with the largest corresponding abnormal case retention rate (i.e., the second screening index) (i.e., the first parameter data point) as one of the advanced candidate parameter data points corresponding to the first type of screening process.

[0064] To further clarify, in this other embodiment, in step S23, processing unit 1 uses the abnormal case retention rates corresponding to the preliminary candidate parameter data corresponding to the first type of screening as the second clustering indicators, and uses the abnormal judgment accuracy rates corresponding to the preliminary candidate parameter data corresponding to the first type of screening as the second screening indicators. Furthermore, in step S24 of this other embodiment, processing unit 1 groups multiple preliminary parameter data corresponding to the first type of screening where the abnormal case retention rates (i.e., the second clustering indicators) are equal. In other words, the abnormal case retention rates (i.e., the second clustering indicators) corresponding to the preliminary parameter data in the same group corresponding to the first type of screening are equal. For each group of parameter data 22, processing unit 1 selects the largest corresponding second screening indicator as the advanced candidate parameter data corresponding to the first type of screening.

[0065] After the processing unit 1 selects the advanced candidate parameter data corresponding to the first type of screening process using the second clustering index and the second screening index, the process proceeds to step S25.

[0066] In step S25, the processing unit 1 treats all the first screening indicators as multiple third clustering indicators, and also treats all the first clustering indicators as multiple third screening indicators. More specifically, in this embodiment, the processing unit 1 uses the anomaly judgment accuracy rate as the third clustering indicator and the anomaly case retention rate as the third screening indicator.

[0067] In step S26, the processing unit 1 divides the parameter data 22 into multiple groups based on the third clustering index, and selects one parameter data 22 from each group as another preliminary candidate parameter data corresponding to the second type of screening process based on the third screening index corresponding to the parameter data 22 in each group. More specifically, in this embodiment, the processing unit 1 groups multiple parameter data 22 whose anomaly detection accuracy (i.e., the third clustering index) is equal to each other into the same group. In other words, the anomaly detection accuracy (i.e., the third clustering index) corresponding to the parameter data 22 in the same group are equal to each other. For the parameter data 22 in the same group, the processing unit 1 selects the parameter data 22 with the highest corresponding abnormal case retention rate from the parameter data 22 in the group as another preliminary candidate parameter data corresponding to the second type of screening process.

[0068] To further clarify, in this other embodiment, in step S25, the processing unit 1 uses the abnormal case retention rate as the third grouping indicator and the abnormality judgment accuracy rate as the third screening indicator. Therefore, in step S26, the processing unit 1 groups multiple parameter data 22 whose abnormal case retention rates are equal, and then uses the parameter data 22 in the group with the highest abnormality judgment accuracy rate as another preliminary candidate parameter data corresponding to the second type of screening process.

[0069] After the processing unit 1 uses the third grouping index and the third screening index to select the preliminary candidate parameter data corresponding to the second type of screening process, the process proceeds to step S27.

[0070] In step S27, the processing unit 1 uses multiple third screening indicators that correspond to the preliminary candidate parameter data corresponding to the second type of screening process as multiple fourth cluster indicators, and uses multiple third cluster indicators that correspond to the preliminary candidate parameter data corresponding to the second type of screening process as multiple fourth screening indicators. More specifically, in this embodiment, the processing unit 1 uses multiple abnormal case retention rates that correspond to the preliminary candidate parameter data corresponding to the second type of screening process as the fourth cluster indicators, and uses multiple abnormal judgment accuracy rates that correspond to the preliminary candidate parameter data corresponding to the second type of screening process as the fourth screening indicators.

[0071] In addition, in this other embodiment, the processing unit 1 uses multiple anomaly judgment accuracy rates corresponding to the preliminary candidate parameter data corresponding to the second type of screening process as multiple fourth cluster indicators, and uses multiple anomaly case retention rates corresponding to the preliminary candidate parameter data corresponding to the second type of screening process as multiple fourth screening indicators.

[0072] In step S28, the processing unit 1 divides the preliminary candidate parameter data corresponding to the second type of screening into multiple groups according to the fourth clustering index. For each group of preliminary candidate parameter data corresponding to the second type of screening, the processing unit 1 selects one preliminary candidate parameter data from the preliminary candidate parameter data of that group as another advanced candidate parameter data corresponding to the second type of screening, based on the fourth screening index corresponding to the preliminary candidate parameter data of that group. More specifically, in this embodiment, the processing unit 1 groups multiple preliminary candidate parameter data whose abnormal case retention rates (i.e., the fourth clustering index) corresponding to the preliminary candidate parameter data of the second type of screening are equal to each other into the same group. In other words, the abnormal case retention rates (i.e., the fourth clustering index) corresponding to the preliminary candidate parameter data of the same group corresponding to the second type of screening are equal to each other. For each group of preliminary candidate parameter data corresponding to the second type of screening process, the processing unit 1 selects the preliminary candidate data with the highest anomaly judgment accuracy from the preliminary candidate data of that group as the other advanced candidate parameter data corresponding to the second type of screening process.

[0073] To further clarify, in this other embodiment, the processing unit 1 groups multiple preliminary candidate parameter data whose anomaly detection accuracy rates (i.e., the fourth clustering index) are equal among the preliminary candidate parameter data corresponding to the second type of screening. In other words, the anomaly detection accuracy rates (i.e., the fourth clustering index) corresponding to the preliminary candidate parameter data in the same group corresponding to the second type of screening are equal. For each group of preliminary candidate parameter data corresponding to the second type of screening, the processing unit 1 selects the preliminary candidate data with the largest anomaly case retention rate from the preliminary candidate data corresponding to the second type of screening as another advanced candidate parameter data corresponding to the second type of screening.

[0074] After the processing unit 1 selects the advanced candidate parameter data corresponding to the second type of screening process, the process proceeds to step S29.

[0075] In step S29, the processing unit 1 determines N most applicable parameter data from the one or more advanced candidate parameter data. Specifically, the processing unit 1 determines whether the abnormal case retention rate corresponding to each advanced parameter data is greater than or equal to a retention rate threshold, and if the determination result is yes, the advanced parameter data is selected as one of the N most applicable parameter data. More specifically, the retention rate threshold is, for example, 95%, but is not limited to this. Continuing the example above, in step S24, the processing unit 1 has already selected the first parameter data as one of the advanced candidate parameter data. The processing unit 1 determines that the abnormal case retention rate (i.e., 100%) of the first parameter data is greater than or equal to the retention rate threshold; therefore, the processing unit 1 selects the first parameter data as one of the N most applicable parameter data.

[0076] After processing unit 1 determines the most suitable parameter data for the N records, the screening process ends.

[0077] The above is an example of how the anti-money laundering characterization parameter analysis system 100 in this embodiment implements the anti-money laundering characterization parameter analysis method.

[0078] It should be noted that steps S01 to S07, S11 to S13, S21 to S29, and the flowcharts in Figures 2 to 4 of this embodiment are merely illustrative examples of one possible implementation of the anti-money laundering characterization parameter analysis method of the present invention. It should be understood that even if steps S01 to S07, S11 to S13, and S21 to S29 are combined, split, or rearranged in order, if the resulting process performs similar functions and yields similar results compared to this embodiment, it still falls within the implementable state of the anti-money laundering characterization parameter analysis method of the present invention. Therefore, steps S01 to S07, S11 to S13, S21 to S29, and the flowcharts in Figures 2 to 4 of this embodiment are not intended to limit the scope of implementation of the present invention.

[0079] In summary, the anti-money laundering characterization parameter analysis system 100 uses the processing unit 1 to use one of the anomaly judgment accuracy rate and the anomaly case retention rate as multiple first cluster indicators, and the other as multiple first screening indicators, to select a parameter data 22 from the parameter data 22 in the group as the preliminary candidate parameter data, and then determines the N most applicable parameter data from the preliminary candidate parameter data, so that the processing unit 1 can improve the accuracy of finding the transaction case data 21 of money laundering or terrorist financing by using the N most applicable parameter data, thus effectively achieving the purpose of the present invention.

[0080] However, the above description is merely an embodiment of the present invention and should not be construed as limiting the scope of the present invention. Any simple equivalent changes and modifications made in accordance with the scope of the patent application and the contents of the patent specification shall still fall within the scope of the patent of the present invention.

[0081] 100: Anti-Money Laundering Characteristic Parameter Analysis System 1: Processing Unit 2: Storage Unit 21: Transaction Case Materials 22: Parameter Data S01~S07: Steps S11~S13: Steps S21~S29: Steps

Claims

1. A money laundering characterization parameter analysis system, comprising: a processing unit; and a storage unit electrically connected to the processing unit, storing multiple sets of parameter data, multiple alert case counts corresponding to the parameter data, and multiple abnormal case counts corresponding to the parameter data, wherein, For each parameter and the corresponding number of alert cases, the number of alert cases indicates the number of alert transaction cases in multiple transaction case data, and each alert transaction case is one of the transaction case data that the computer determines has an abnormal risk based on the corresponding parameter. For each parameter and the corresponding number of abnormal cases, the number of abnormal cases indicates the number of abnormal transaction case data in the transaction case data, and each abnormal transaction case is one of the alert transaction case data corresponding to the parameter and is manually determined to be abnormal. The processing unit selects the largest abnormal case from the abnormal case numbers as a reference abnormality number. The processing unit calculates an abnormality judgment accuracy rate corresponding to the parameter data based on the proportional relationship between the abnormal case number corresponding to each parameter and the number of alert cases corresponding to the parameter. The processing unit calculates an abnormality case retention rate corresponding to the parameter data based on the proportional relationship between the abnormal case number corresponding to each parameter and the reference abnormality number. The processing unit performs a filtering procedure on the parameter data to select N most applicable parameter data from the parameter data, and generates and outputs a parameter filtering result containing the N most applicable parameter data, where N is an integer greater than or equal to 1. The processing unit performs the filtering procedure in the following ways: using one of the two types of data, namely the anomaly judgment accuracy rate and the anomaly case retention rate, as multiple first clustering indicators, and using the other type of data as multiple first filtering indicators; dividing the parameter data into multiple groups according to the first clustering indicators, and selecting one parameter data from the parameter data in each group as a preliminary candidate parameter data according to the first filtering indicators corresponding to the parameter data in each group; and determining the N most applicable parameter data from the preliminary candidate parameter data.

2. The anti-money laundering characterization parameter analysis system as described in claim 1, wherein, The processing unit determines the N most applicable parameter data from the preliminary candidate parameter data in the following manner: It uses multiple first screening indicators corresponding to the preliminary candidate parameter data as multiple second clustering indicators, and uses multiple first clustering indicators corresponding to the preliminary candidate parameter data as multiple second screening indicators; it divides the preliminary candidate parameter data into multiple groups according to the second clustering indicators, and selects one preliminary candidate parameter data from the preliminary candidate parameter data in each group as an advanced candidate parameter data according to the second screening indicators corresponding to the preliminary candidate parameter data in each group; and it determines the N most applicable parameter data from the one or more advanced candidate parameter data.

3. The anti-money laundering characterization parameter analysis system as described in claim 2, wherein, The processing unit performs the screening procedure by further comprising the following steps before determining the N most applicable parameter data from the one or more advanced candidate parameter data: treating the first screening indicators as multiple third grouping indicators, and treating the first grouping indicators as multiple third screening indicators; dividing the parameter data into multiple groups according to the third grouping indicators; for each group of parameter data, selecting one parameter data from the group of parameter data as another preliminary candidate parameter data according to the third screening indicators corresponding to the parameter data in that group; treating multiple third screening indicators corresponding to the preliminary candidate parameter data as multiple fourth grouping indicators, and treating multiple third grouping indicators corresponding to the preliminary candidate parameter data as multiple fourth screening indicators; Based on the fourth grouping index, the preliminary candidate parameter data are divided into multiple groups. For each group of preliminary candidate parameter data, based on the fourth screening index corresponding to the preliminary candidate parameter data of the group, one preliminary candidate parameter data is selected from the preliminary candidate parameter data of the group as another advanced candidate parameter data.

4. The anti-money laundering characterization parameter analysis system as described in any one of claims 2 and 3, wherein, The processing unit determines the N most applicable parameter data from the one or more advanced candidate parameter data in the following way: it determines whether the retention rate of the abnormal case corresponding to each advanced parameter data is greater than or equal to a retention rate threshold value, and if the determination result is yes, the advanced parameter data is regarded as one of the N most applicable parameter data.

5. The anti-money laundering characterization parameter analysis system as described in claim 1, wherein, The storage unit also stores transaction case data, each transaction case data including a transaction amount and an attribute tag, and each parameter data including multiple amount threshold values ​​corresponding to a case attribute feature; the processing unit performs a check process on each transaction case data according to each parameter data, wherein the processing unit performs the check process by using the attribute tag of the transaction case data as the attribute threshold value corresponding to the case attribute feature in the parameter data that matches the attribute tag as a matching amount threshold value, comparing the transaction amount of the transaction case data with the matching amount threshold value, and when it is determined that the transaction amount is greater than or equal to the matching amount threshold value, the transaction case data is treated as a warning transaction case data corresponding to the parameter data; the processing unit counts the number of warning transaction case data corresponding to each parameter data to obtain the number of warning cases corresponding to the parameter data, and stores the number of warning cases in the storage unit; For each parameter data, the processing unit treats each warning transaction case with an abnormal annotation in the warning transaction case data corresponding to the parameter data as an abnormal transaction case data corresponding to the parameter data, counts the number of abnormal transaction case data corresponding to each parameter data to obtain the abnormal case quantity corresponding to the parameter data, and stores the abnormal case quantity in the storage unit.

6. A method for analyzing anti-money laundering characterization parameters, implemented by an anti-money laundering characterization parameter analysis system, wherein the system stores multiple sets of parameter data, multiple alert case volumes corresponding to the parameter data, and multiple abnormal case volumes corresponding to the parameter data, wherein... For each parameter and the corresponding number of alert cases, the number of alert cases indicates the number of alert transaction cases in multiple transaction case data, and each alert transaction case is one of those transaction case data that the computer determines has an abnormal risk based on the corresponding parameter. For each parameter and the corresponding number of abnormal cases, the number of abnormal cases indicates the number of abnormal transaction case data in the transaction case data, and each abnormal transaction case is one of those transaction case data that matches the parameter and is manually determined to be abnormal; the anti-money laundering characterization parameter analysis method includes: (A) The anti-money laundering characterization parameter analysis system selects the largest one from the abnormal case quantities as a reference abnormal quantity; (B) The anti-money laundering characterization parameter analysis system calculates an abnormality judgment accuracy corresponding to the parameter data based on the proportional relationship between the abnormal case quantity corresponding to each parameter and the alert case quantity corresponding to the parameter; (C) The anti-money laundering characterization parameter analysis system calculates an abnormal case retention rate corresponding to the parameter data based on the proportional relationship between the number of abnormal cases corresponding to each parameter data and the reference abnormality amount; (D) The anti-money laundering characterization parameter analysis system performs a screening procedure on the parameter data to select N most applicable parameter data from the parameter data, and generates and outputs a parameter screening result containing the N most applicable parameter data, where N is an integer greater than or equal to 1. The anti-money laundering characterization parameter analysis system performs the screening procedure in the following ways: (D1) Using one of the two data, the abnormality judgment accuracy rate and the abnormal case retention rate, as multiple first cluster indicators, and using the other data as multiple first screening indicators; (D2) Dividing the parameter data into multiple groups according to the first cluster indicators, and selecting one parameter data from the parameter data in each group as a preliminary candidate parameter data according to the first screening indicators corresponding to the parameter data in each group; and (D3) The most suitable N parameter data are determined from these preliminary candidate parameter data.

7. The method for analyzing anti-money laundering characterization parameters as described in claim 6, wherein, In step (D3), the method by which the anti-money laundering characterization parameter analysis system determines the N most applicable parameter data from the preliminary candidate parameter data includes: (D31) using multiple first screening indicators corresponding to the preliminary candidate parameter data as multiple second grouping indicators, and using multiple first grouping indicators corresponding to the preliminary candidate parameter data as multiple second screening indicators; (D32) dividing the preliminary candidate parameter data into multiple groups according to the second grouping indicators, and selecting one preliminary candidate parameter data from the preliminary candidate parameter data of each group as an advanced candidate parameter data according to the second screening indicators corresponding to the preliminary candidate parameter data of each group; and (D33) determining the N most applicable parameter data from the one or more advanced candidate parameter data.

8. The method for analyzing anti-money laundering characterization parameters as described in claim 7, wherein, In step (D3), the method by which the anti-money laundering characterization parameter analysis system performs the screening procedure further includes, before determining the N most applicable parameter data from the one or more advanced candidate parameter data: (D34) treating the first screening indicators as multiple third grouping indicators, and treating the first grouping indicators as multiple third screening indicators; (D35) dividing the parameter data into multiple groups according to the third grouping indicators, and for the parameter data in each group, selecting one parameter data from the parameter data in that group as another preliminary candidate parameter data according to the third screening indicators corresponding to the parameter data in that group; (D36) The third screening indicators corresponding to the preliminary candidate parameter data are respectively used as a plurality of fourth grouping indicators, and the third grouping indicators corresponding to the preliminary candidate parameter data are respectively used as a plurality of fourth screening indicators; and (D37) The preliminary candidate parameter data are divided into a plurality of groups according to the fourth grouping indicators, and for the preliminary candidate parameter data of each group, the anti-money laundering characterization parameter analysis system selects one preliminary candidate parameter data from the preliminary candidate parameter data of the group as another advanced candidate parameter data according to the fourth screening indicators corresponding to the preliminary candidate parameter data of the group.

9. The method for analyzing anti-money laundering characterization parameters as described in claim 7 or 8, wherein, In step (D33), the method by which the anti-money laundering characterization parameter analysis system determines the N most applicable parameter data from one or more advanced candidate parameter data includes: determining whether the retention rate of the abnormal case corresponding to each advanced parameter data is greater than or equal to a retention rate threshold value, and if the determination result is yes, the advanced parameter data is regarded as one of the N most applicable parameter data.

10. The anti-money laundering characterization parameter analysis method as described in claim 6, wherein the anti-money laundering characterization parameter analysis system further stores transaction case data, each transaction case data including a transaction amount and an attribute tag, each parameter data including multiple amount threshold values ​​corresponding to a case attribute feature; the anti-money laundering characterization parameter analysis method further includes, prior to step (A): (E) the anti-money laundering characterization parameter analysis system performing a verification process on each transaction case data based on each parameter data, wherein, The method by which the anti-money laundering characterization parameter analysis system performs the verification process includes: (E1) using the attribute marker of the transaction case data as the amount threshold value corresponding to the case attribute feature that matches the attribute marker in the parameter data as a matching amount threshold value; (E2) comparing the transaction amount of the transaction case data with the matching amount threshold value; (E3) when it is determined that the transaction amount is greater than or equal to the matching amount threshold value, treating the transaction case data as a warning transaction case data corresponding to the parameter data; (F) the anti-money laundering characterization parameter analysis system counts the number of warning transaction case data corresponding to each parameter data to obtain the number of warning cases corresponding to the parameter data, and stores the number of warning cases; and (G) for each parameter data, the anti-money laundering characterization parameter analysis system treats each warning transaction case data with an abnormal annotation in the warning transaction case data corresponding to the parameter data as an abnormal transaction case data corresponding to the parameter data, counts the number of abnormal transaction case data corresponding to each parameter data to obtain the number of abnormal cases corresponding to the parameter data, and stores the number of abnormal cases.

Images