TWI767161B - Method for evaluating trading partner - Google Patents

Abstract

The present invention provides a method for evaluating a trading partner, which includes: collecting a score and multiple association attribute values of each of a plurality of trading partners, and accordingly establishing an association matrix; finding a plurality of visiting paths between the first trading partner and the second trading partner based on the association matrix; estimating the path cost of each visiting path based on the score of each second trading partner; estimating a trading risk score of the first trading partner based on the path cost of each visiting path.

Description

Approach to Evaluate Transaction Objects

The present invention relates to a risk assessment mechanism, and in particular to a method for assessing transaction objects.

In view of the fact that when evaluating the transaction object in the past, the company's risk was only evaluated from a single aspect such as news search and official website. When the transaction object is a newly registered company without historical information and background, the company should judge whether the transaction object is a Reliable partners have practical difficulties. Many companies currently have a relationship between major companies and other companies, and most of them have a system of referring to a whitelist or a blacklist. It may be difficult to detect with the existing credit investigation methods by setting up a new company or finding another related company to seek cooperation opportunities with the company in a roundabout way.

In view of this, the present invention provides a method for evaluating a transaction object, which can be used to solve the above-mentioned technical problems.

The present invention provides a method for evaluating transaction objects, comprising: collecting a score and a plurality of associated attribute values of a plurality of transaction objects, wherein the transaction objects include a first transaction object and a plurality of second transaction objects; based on the transaction objects The individual aforementioned associated attribute values establish an association matrix, wherein the association matrix records a correlation coefficient between each transaction object and the aforementioned associated attribute value; based on the association matrix, find out a plurality of the relationship between the first transaction object and the aforementioned second transaction object. A visit path, wherein each visit path includes a plurality of nodes connected in series, and an initial node in the aforementioned nodes corresponds to the first transaction object; a path cost of each visit path is estimated based on the score of each second transaction object; based on each visit The path cost of the path estimates a transaction risk score for the first transaction object.

In order to make the above-mentioned features and advantages of the present invention more obvious and easy to understand, the following embodiments are given and described in detail with the accompanying drawings as follows.

Roughly speaking, the present invention hopes to incorporate the experience of cooperation with companies in the past, and more specifically, to define the risk level in a formalized way for the relationship between transaction objects, to provide a more effective and multi-faceted analysis result, and to develop an applicable The risk level assessment of all transaction objects will be further explained below.

In order to facilitate the understanding of the concept of the present invention, an actual situation is supplemented below for description. In the context of the present invention, it is assumed that the transaction object 02 has a business relationship with the company A implementing the method of the present invention, the construction quality is not good, and there is a financial dispute with the supplier, so the transaction object 02 is assigned a lower score. . Since company A tends to avoid cooperating with high-risk counterparties, counterparty 02 turns to establish a new company B to participate in the bidding of company A, and this new company B is counterparty 05. The transaction object 05 is a new transaction object in the market. According to the general assessment standard, it can obtain a medium rating score, but the management personnel are related to the transaction object 02, which means that the transaction object 05 may become a high-risk transaction object in the future, resulting in The company suffered losses. The above situation often occurs in practice, and the method of the present invention can obtain a risk rating score for reference, so as to achieve the effect of gatekeeping for the enterprise.

Please refer to FIG. 1 , which is a flowchart of a method for evaluating a transaction object according to an embodiment of the present invention. In different embodiments, the method of the present invention may be executed by a specific computer server, workstation, or smart device to implement the steps in FIG. 1 based on the operations of Company A personnel, which are described in detail below.

First, in step S110, individual scores and multiple associated attribute values of a plurality of transaction objects may be collected, wherein the foregoing transaction objects include a first transaction object and a plurality of second transaction objects. In the embodiment of the present invention, the first transaction object is, for example, a transaction object to be evaluated (ie, transaction object 05 ) input by the user, and the second transaction object is, for example, available to assist in evaluating the first transaction object Other transaction objects (eg transaction object 02).

In one embodiment, it is assumed that the present invention considers transaction objects 00, 01, 02, 03, 04, 05 and associated attribute values a, b, c, and d. In different embodiments, the associated attribute values a~d may have different implementation aspects, such as geographic region, directors and supervisors, contact number, representative, shareholding ratio, etc., but are not limited thereto. Based on this, the present invention can, for example, by accessing/searching historical data, public information, Ministry of Finance registration data, news or crawling in the database, etc., after integrating various pipeline resources, each transaction object 00-05 can be associated with the corresponding one. The attribute values a~d are assembled into an association table as shown in Figure 2.

Please refer to FIG. 2 , which is an association table showing each transaction object according to an embodiment of the present invention. It can be seen from Figure 2 that transaction objects 00, 01, and 04 all have associated attribute values d (for example, they are located in a specific county and city); transaction objects 04 and 02 both have associated attribute values a (eg, have one or more identical directors and supervisors); both transaction objects 02 and 05 have associated attribute value c (for example, have the same contact number); only transaction object 03 has associated attribute value b.

In addition, the score of each transaction object 00-05 is, for example, the transaction risk score previously estimated for each transaction object 00-05 by the method of the present invention, but it is not limited to this. In one embodiment, the scores of each transaction object 00-05 may be exemplified in Table 1 below. In addition, in the context of the present invention, since the transaction object 05 corresponds to the new company B, there is currently no relevant score. Trading partners score 00 60 01 40 02 20 03 80 04 70 05 unknown Table 1

Afterwards, in step S120, an association matrix may be established based on the individual associated attribute values of transaction objects 00-05. In one embodiment, the dimension of the above-mentioned correlation matrix (represented by R[X, Y]) is, for example, MxN, where M is the number of transaction objects 00-05 (ie, 6), and N is the correlation attribute value a-d number (ie, 4). Moreover, if the i-th transaction object in the transaction objects 00-05 has the j-th associated attribute value in the associated attribute values a-d, the present invention can set the element of the i-th column and the j-th row in the association matrix as a Correlation coefficient (for example, 1).

Please refer to FIG. 3 , which is a schematic diagram of an association matrix according to an embodiment of the present invention. In this embodiment, the set of transaction objects 00~05 can be expressed as X={X ₀₀ , X ₀₁ , X ₀₂ , X ₀₃ , X ₀₄ , X ₀₆ }, and the set of associated attribute values a~d can be expressed as Y={Y _a , Y _b , Y _c , Y _d }, but not limited thereto.

According to Fig. 2, since the transaction object 00 (ie, the first transaction object) has an associated attribute value d (ie, the fourth associated attribute value), the elements in the first column and the fourth row of the association matrix 300 in Fig. 3 (which can be expressed as R[X ₀₀ , Y _d ]) can be set to 1 accordingly. Similarly, since the transaction object 04 (ie, the fifth transaction object) also has an associated attribute value d, the element in the fifth column and the fourth row of the association matrix 300 in FIG. 3 (ie, R[X ₀₄ , Y _d ] ) can be set to 1 accordingly. Based on the above teachings, those skilled in the art should understand the meanings of other elements in the correlation matrix 300 , and details are not described herein.

Next, in step S130 , based on the correlation matrix 300 , a plurality of visiting paths (or intelligible paths) between the first transaction object (ie, transaction object 05 ) and the second transaction object (ie, transaction objects 00 to 04 ) can be found. is the associated path).

Specifically, from R[X ₀₆ , Y _c ]=1 and R[X ₀₂ , Y _c ]=1 in FIG. 3 , it can be known that the newly added transaction object 05 and the transaction object 02 are directly associated through the associated attribute value c , which means that the transaction objects 02 and 05 have the same association attribute value c, and the association hop number is 1, indicating that the association between the two is very high. Similarly, it can be seen from FIG. 3 that the transaction object 04 is associated with the transaction object 02 through the associated attribute value a, and the transaction object 02 is associated with the transaction object 06 through the associated attribute value c, indicating that the transaction object 04 is associated with the transaction Object 06 is associated through transaction object 02, which is an indirect association, and the number of hops is 2. The more hops in the indirect association, the weaker the association between the two transaction objects, but there is still a correlation.

In one embodiment, in the process of generating the correlation matrix 300 , a visiting path between the first transaction object (ie, transaction object 05 ) and the second transaction object (ie, transaction objects 00 to 04 ) is also formed.

Please refer to FIG. 4 , which is a schematic diagram of a visiting route according to an embodiment of the present invention. In this embodiment, according to the correlation matrix 300 , visiting paths 410 and 420 can be formed between the first transaction object (ie, transaction object 05 ) and the second transaction object (ie, transaction objects 00 to 04 ).

Roughly speaking, for a specific visit path in the visit paths 410 and 420, it may include T specific nodes connected in series, each specific node corresponds to one of the transaction objects 00-05, and the No. 1 specific node corresponds to the first transaction object (ie, transaction object 05). In addition, the k-th specific node and the k-1-th specific node in the aforementioned specific nodes both have the first associated attribute value in the aforementioned associated attribute values, wherein k is an integer greater than 1 and less than T; Both the k+1 specific nodes and the kth specific node have the second associated attribute value among the aforementioned associated attribute values, and the first associated attribute value is different from the second associated attribute value.

Taking the visit path 410 as an example, the first specific node therein corresponds to the transaction object 05 . As can be seen from FIG. 3 , since both the transaction object 02 and the transaction object 05 have the associated attribute value c, the second specific node in the visit path 410 corresponds to the transaction object 02 . In addition, since both the transaction object 04 and the transaction object 02 have the associated attribute value a, the third specific node in the visit path 410 corresponds to the transaction object 04 . Similarly, since both the transaction object 00 and the transaction object 04 have the associated attribute value d, the fourth specific node in the visit path 410 corresponds to the transaction object 00 .

In addition, it can be seen from the visit path 410 that there are k-1 hops between the kth specific node and the 1st specific node. For example, there is 1 hop between the 2nd (k=2) specific node (corresponding to transaction object 02) and the 1st specific node; the 3rd (k=3) specific node (corresponding to transaction object 04) and the There are 2 hops between the 1st specific node; there are 3 hops between the 4th (k=4) specific node (corresponding to transaction object 00) and the 1st specific node.

Taking the visit path 420 as an example again, the first specific node therein corresponds to the transaction object 05 . As can be seen from FIG. 3 , since both the transaction object 02 and the transaction object 05 have the associated attribute value c, the second specific node in the visit path 420 corresponds to the transaction object 02 . In addition, since both the transaction object 04 and the transaction object 02 have the associated attribute value a, the third specific node in the visit path 420 corresponds to the transaction object 04 . Similarly, since both the transaction object 01 and the transaction object 04 have the associated attribute value b, the fourth specific node in the visit path 420 corresponds to the transaction object 01 .

In addition, it can also be seen from the visit path 420 that there is a jump between the second (k=2) specific node (corresponding to transaction object 02) and the first specific node; the third (k=3) specific node There are 2 hops between the node (corresponding to transaction object 04) and the 1st specific node; there are 3 hops between the 4th (k=4) specific node (corresponding to transaction object 01) and the 1st specific node .

After the visiting routes 410 and 420 are obtained, in step S140 , the route cost of each visiting route 410 and 420 may be estimated based on the score of each second transaction object.

In one embodiment, for a specific visit path in the visit paths 410 and 420, the present invention can estimate the weight score of each specific node in the specific visit path, and calculate the average score of the weight score of each specific node accordingly. , as the path cost for a particular visit path.

Roughly speaking, for the kth specific node in a specific visit path, the present invention can obtain a weight value corresponding to the k-1 hops, wherein the weight value is negatively related to the k-1 hops ( That is, the more hops a particular node has, the lower the weight value). Afterwards, the present invention can obtain the score of the second transaction object corresponding to the kth specific node, and modify it into the weight score of the kth specific node according to the weight value. Then, the present invention can obtain the average score of the weight scores of each of the k-th specific nodes as the path cost of the specific visiting path.

Taking the visit path 410 as an example, for the second specific node (corresponding to the transaction object 02), the present invention can obtain the weight value (for example, 1) corresponding to one jump, and then the present invention can obtain the second specific node. The score (ie, 20) of the second transaction object (ie, transaction object 02) corresponding to the node, and according to the above weight value (ie, 1), it is revised to the weight score of the second specific node. In one embodiment, the weight score of the second specific node is, for example, 20 (ie, 20×1).

For the third specific node (corresponding to the transaction object 04) in the visiting path 410, the present invention can obtain the weight value (for example, 0.9) corresponding to two hops, and then the present invention can obtain the third specific node corresponding to the third specific node. The score (ie, 70) of the second transaction object (ie, the transaction object 04), and according to the above-mentioned weight value (ie, 0.9), it is corrected to the weight score of the third specific node. In one embodiment, the weight score of the third specific node is, for example, 63 (ie, 70×0.9).

For the 4th specific node (corresponding to the transaction object 00) in the visiting path 410, the present invention can obtain the weight value corresponding to 3 hops (for example, 0.8), and then the present invention can obtain the 4th specific node corresponding to the 4th specific node. Second, the score (ie, 60) of the transaction object (ie, the transaction object 00), and according to the above weight value (ie, 0.8), it is revised to the weight score of the fourth specific node. In one embodiment, the weight score of the 4th specific node is, for example, 48 (ie, 60x0.8).

Afterwards, the present invention can obtain the average score of the weight scores of each specific node in the visited path 410 as the path cost of the visited path 410 . In one embodiment, the path cost of visiting path 410 is, for example, 43.7 (ie, (20+63+48)/3).

Taking the visit path 420 as an example again, for the second specific node (corresponding to the transaction object 02), the present invention can obtain the weight value (for example, 1) corresponding to one jump, and then the present invention can obtain the second node The score (ie, 20) of the second transaction object (ie, the transaction object 02) corresponding to the specific node, and it is corrected to the weight score of the second specific node according to the above weight value (ie, 1). In one embodiment, the weight score of the second specific node is, for example, 20 (ie, 20×1).

For the third specific node (corresponding to the transaction object 04) in the visiting path 420, the present invention can obtain the weight value (for example, 0.9) corresponding to two hops, and then the present invention can obtain the third specific node corresponding to the third specific node. The score (ie, 70) of the second transaction object (ie, the transaction object 04), and according to the above-mentioned weight value (ie, 0.9), it is corrected to the weight score of the third specific node. In one embodiment, the weight score of the third specific node is, for example, 63 (ie, 70×0.9).

For the 4th specific node (corresponding to the transaction object 01) in the visiting path 420, the present invention can obtain the weight value corresponding to 3 hops (for example, 0.8), and then the present invention can obtain the 4th specific node corresponding to the 4th specific node. The score (ie, 40) of the second transaction object (ie, the transaction object 01), and according to the above weight value (ie, 0.8), it is revised to the weight score of the 4th specific node. In one embodiment, the weight score of the 4th specific node is, for example, 32 (ie, 40×0.8).

Afterwards, the present invention can obtain the average score of the weight scores of each specific node in the visiting path 420 as the path cost of the visiting path 420 . In one embodiment, the path cost of visiting path 420 is, for example, 38.3 (ie, (20+63+32)/3).

After obtaining the path cost (ie, 43.7 and 38.3) of each visit path 410, 420, in step S150, the transaction of the first transaction object (ie, transaction object 05) may be estimated based on the path cost of each visit path 410, 420 Risk Score. In one embodiment, the present invention can obtain the average cost of the path costs of each visiting path 410, 420 as the transaction risk score of the first transaction object (ie, transaction object 05), for example, 41 (ie, (43.7+ 38.3)/2). In this case, Table 1 can be updated as exemplified in Table 2 below. Trading partners score 00 60 01 40 02 20 03 80 04 70 05 41 Table 2

After obtaining the transaction risk score of the transaction object 05, it can be provided to the internal personnel of the company A as a reference, which becomes an important basis for whether to cooperate with the transaction object 05. Several transaction objects proposed in this case are more or less related to transaction object 02, so they also indirectly affect their respective ratings. After the new transaction object 05 calculates the transaction risk score through the method of the present invention, it can obtain the rating score related to each transaction object 00~04 through direct and indirect association, and finally these scores will be archived and fed back to enrich the association database .

The rated transaction objects can be brought into the method of the present invention once to obtain a new score (ie, a transaction risk score) if the conditions are adjusted later, or when the survey is re-investigated. However, in order to prevent the transaction object from deliberately changing the conditions to obtain a higher score, the present invention also proposes a related mechanism for modifying the risk assessment score.

表示）。之後，本發明可依據交易對象05的評分（以

表示）修正交易對象05的交易風險分數。Taking the transaction object 05 as an example, suppose that the personnel of company A want to evaluate the transaction object 05 again after using the method of the present invention to perform steps S110 to S150 on it for the first time to obtain the score (ie, 41) recorded in Table 2. Then, the transaction risk score of transaction object 05 can be obtained based on steps S110 to S150 again (with

express). After that, the present invention can be based on the score of the transaction object 05 (with

indicates) to correct the transaction risk score of transaction object 05.

，其中

與

分別為

與

對應的權重值，

與

的總和為1。In one embodiment, the revised transaction risk score of transaction object 05 can be characterized as:

,in

and

respectively

and

The corresponding weight value,

and

The sum is 1.

與

可依設計者的需求而選用所需的數值。在一實施例中，為提升

的比重（即，較重視當次估計而得的交易風險分數），

可選定為小於

的數值，但可不限於此。舉例而言，假設

為50，

為0.4，

為0.6，則交易對象05的修正後交易風險分數可估計為46.4（即，((0.4x41)+(0.6x50))/(0.6+0.4)）。In various embodiments,

and

The desired value can be selected according to the needs of the designer. In one embodiment, to enhance

(that is, placing more emphasis on the transaction risk score obtained from the current estimate),

can be selected to be less than

value, but not limited to this. For example, suppose

is 50,

is 0.4,

is 0.6, then the revised transaction risk score of transaction object 05 can be estimated to be 46.4 (ie, ((0.4x41)+(0.6x50))/(0.6+0.4)).

Correspondingly, the internal personnel of company A can use the revised transaction risk score (ie, 46.4) of the transaction object 05 as the basis for the consideration of the partnership, and incorporate it into the related database to achieve the effect of retraining.

To sum up, the method of the present invention can analyze the transaction risk rating of the transaction object (customer or manufacturer) to be evaluated based on the degree of correlation between companies or enterprises and historical transaction records. For example, the present invention can obtain a plurality of transaction objects, and then establish the relationship between these transaction objects according to the associated attribute value, and generate an association matrix. From the correlation matrix, one or more visiting paths can be formed, establish a risk correlation matrix and obtain the scores between nodes on the path, and then calculate the transaction risk rating through the formula, and finally reflect the transaction risk rating back to the transaction object. Score set for risk association retraining.

The present invention can discover the relationship between companies or enterprises, and finally generate a transaction object risk rating by visiting multiple nodes. The rating can effectively reflect the potential risks of future trading behavior. Moreover, the present invention can detect suspicious transaction objects. After entering the transaction object to be evaluated, the transaction risk rating can be calculated through the relevant process of this patent, and fed back to the score set of the transaction object.

As can be seen from the above, the present invention has at least the following characteristics: (1) The relationship between companies or enterprises can be discovered, and the associated evaluation score can be generated for evaluating transaction objects. The evaluation can effectively reflect future transaction behaviors (2) It can effectively conduct risk assessment for the first-time cooperation or new market participants, even if there is not enough information to refer to; (3) It has the ability to give feedback and learn. Each time the risk rating operation is performed, the risk value of the transaction object will be incorporated into the historical cooperation experience as a reference for obtaining the risk correlation score. Using the results of each evaluation execution as empirical learning can reduce the possibility of taking too much risk on huge transactions. In addition, recording each experience with the same standard can also make the relationship between high-quality transaction objects and enterprises have a more positive cycle; (4) It can reduce the unknown risk of the company undertaking special bidding cases.

Although the present invention has been disclosed above by the embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the technical field can make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, The protection scope of the present invention shall be determined by the scope of the appended patent application.

S110~S150: Steps 00~05: Transaction object a~d: associated attribute value 300: Correlation Matrix 410, 420: Visiting path

FIG. 1 is a flowchart of a method for evaluating a transaction object according to an embodiment of the present invention. FIG. 2 is an association table illustrating transaction objects according to an embodiment of the present invention. FIG. 3 is a schematic diagram of an association matrix according to an embodiment of the present invention. FIG. 4 is a schematic diagram of a visiting route according to an embodiment of the present invention.

S110~S150: Steps

Claims (5)

A method for evaluating transaction objects, suitable for an electronic device, comprising: collecting, by the electronic device, a score and a plurality of associated attribute values of a plurality of transaction objects, wherein the transaction objects include a first transaction object and a plurality of first transaction objects. Two transaction objects; an association matrix is established by the electronic device based on the respective associated attribute values of the transaction objects, wherein the association matrix records an association coefficient between each transaction object and the associated attribute values, wherein the association The dimension of the matrix is MxN, wherein M is the number of the transaction objects, and N is the number of the associated attribute values, and the step of establishing the association matrix based on the individual associated attribute values of the transaction objects includes: responding to Determine that the i-th transaction object among the transaction objects has the j-th associated attribute value among the associated attribute values, and set the element of the i-th column and the j-th row in the association matrix as the association coefficient; by the electronic device Find out a plurality of visiting paths between the first transaction object and the second transaction objects based on the correlation matrix, wherein each of the visiting paths includes a plurality of nodes connected in series, and an initial node of the nodes corresponds to the first transaction object; the electronic device estimates a path cost of each visiting path based on the score of each second transaction object; the electronic device estimates the first transaction object's cost based on the path cost of each visiting path a transaction risk score, wherein the visit paths include a specific visit path, the specific visit path includes T specific nodes connected in series, each of the specific nodes corresponds to one of the transaction objects, and the first one of the specific nodes 1 specific node pair Corresponding to the first transaction object, wherein: the k-th specific node and the k-1-th specific node in the specific nodes both have a first associated attribute value among the associated attribute values, wherein k is greater than 1 an integer less than T; the k+1 th specific node and the k th specific node in the specific nodes both have a second related attribute value among the related attribute values, and the first related attribute value is different from the The second associated attribute value, wherein k-1 hops exist between the kth specific node and the first specific node, and the path of each visit path is estimated based on the score of each second transaction object The step of costing includes: for the kth specific node, obtaining a weight value corresponding to the k-1 jumps, wherein the weight value is negatively related to the k-1 jumps; obtaining the The score of the second transaction object corresponding to the k specific nodes is corrected to a weight score of the kth specific node according to the weight value; the score of the weight score of each of the kth specific nodes is obtained. A score average to serve as the path cost for that particular visit path. The method of claim 1, wherein the step of estimating the transaction risk score of the first transaction object based on the path cost of each of the visiting paths comprises: obtaining a cost average of the path costs of each of the visiting paths value as the transaction risk score of the first transaction object. The method as described in item 1 of the scope of the application, further comprising: The transaction risk score of the first transaction object is modified according to the score of the first transaction object. The method described in item 3 of the patent application scope, wherein the revised transaction risk score is characterized as:

, where _SM is the _{score of the first transaction object, SN} is the transaction risk score of the first transaction object, and the sum of α and β is 1. The method of claim 4, wherein α is less than β .

Images