KR20110117475A - Apparatus and method for inferring trust and reputation in web-based social network - Google Patents

Abstract

An apparatus and method for trust and reputation inference in a web-based social network are disclosed. The path reliability inference unit infers the path reliability score of the trust object for the trust entity based on the trust between intermediate nodes located on the path from the trust entity that requested the trust score of the trust object to the trust object in the web-based social network. do. The confidence summing unit weights and sums each of the inferred path reliability scores for each of the plurality of paths from the trusting subject to the trusting object, and calculates the final reliability score of the trusting object for the trusting subject, and assigns each path reliability score. The weight to be set is set to a value inversely proportional to the difference from the average of the plurality of path reliability scores. The reputation inference unit infers the reputation score of the trust object in the social network by summing the final reliability scores calculated for each node having a path connected to the trust object. According to the present invention, when the trust subject and the trust object are connected to each other through a plurality of paths, the final reliability score of the trust object is calculated by weighting the path reliability scores of the trust objects inferred for each path, thereby obtaining the plurality of trust objects. It is possible to increase the contribution of the values close to the mean of the path reliability score and improve the accuracy of the final reliability score.

Description

Apparatus and method for inferring trust and reputation in web-based social network}

The present invention relates to an apparatus and method for trust and reputation inference in a web-based social network. More particularly, the present invention relates to a reliability and reputation in a network in which a plurality of members are connected to each other. An apparatus and method for formulating and quantifying a plate.

With the widespread use of the Internet, the online business has grown dramatically over the last few years. Traditional off-line businesses have become increasingly available online, and numerous new business opportunities have been introduced into the online environment. For example, an online transaction that allows the exchange of goods and services entirely computerized can provide the benefits of cost savings and increased convenience. Technical issues related to online transactions, such as security and network availability, have improved and reached a near-stable level, while sociological aspects such as trust and reputation still require extensive research. In an online trading environment, trust relationships are difficult to achieve because there is no physical contact or interaction between the people involved in such a transaction.

Existing online trading systems provide information such as the user's personal information, transaction history and comments or ratings of previous consumers to help the online user make decisions regarding reliability. However, most of these systems only provide intuitive access to reliability and reputation, so the information provided is incomplete, ambiguous and unreliable. Therefore, a systematic approach is needed to formulate the sociological concept of trustworthiness and transform it into quantitative computational information for reliable online transactions.

Reliability is an essential element of building relationships between individuals or organizations, and reputation is the public opinion or expectation of an individual based on the individual's behavior. Reliability and reputation influence all behaviors and skills, including interactions between people, and serve as a barometer to estimate the creditworthiness of potential parties. Many studies have now been conducted to formulate reliability and reputation through computational models.

One of the existing researches is an algorithm that aggregates and infers reputation and reliability on semantic web-based social networks. In this algorithm, the reliability between two individuals who are not directly connected in the network is inferred based on the locally calculated intermediate node's reliability rank. We also proposed a quantitative model that calculates reputation by summing up the reliability scores of the searched paths. However, this method has the disadvantage that accuracy can be inferred because the reliability inference results can vary according to many paths that can exist between the two individuals.

The technical problem to be achieved by the present invention is to plot the relationship between the members in the network to increase the accuracy in consideration of the reliability between the various paths to the member that is the object of reliability inference and the members located on the path, To provide a reliability and reputation inference device and method in a web-based social network that can reduce the computational load.

Another technical problem to be achieved by the present invention is to plot the relationship between members in the network to increase the accuracy in consideration of the reliability between the various paths to the member that is the object of reliability inference and the members located on the path, The present invention provides a computer-readable recording medium that records a program for executing a method of inferring reliability and reputation in a web-based social network that can reduce a computational load.

In order to achieve the above technical problem, a reliability and reputation inference apparatus in a web-based social network according to the present invention is for evaluating online transaction credit of a trust object selected from a plurality of nodes constituting a web-based social network. And inferring a path reliability score of the trusted object for the trusted object based on the reliability between an intermediate node located on a path from the trusted subject requesting the trusted score of the trusted object to the trusted object in the social network. Path reliability inference unit; Each of the path reliability scores inferred for each of the plurality of paths from the trusted subject to the trusted object is weighted and summed to calculate a final reliability score of the trusted object for the trusted subject, wherein the respective path reliability scores are calculated. A reliability adder configured to set a weight assigned to the value in inverse proportion to a difference from an average of the plurality of path reliability scores; And a reputation inference unit for inferring a reputation score of the trust object in the social network by summing final final confidence scores calculated for nodes having a path connected to the trust object.

In order to achieve the above technical problem, the reliability and reputation inference method in a web-based social network according to the present invention is for evaluating the online transaction credit of a trust object selected from a plurality of nodes constituting the web-based social network. And inferring a path reliability score of the trusted object for the trusted object based on the reliability between an intermediate node located on a path from the trusted subject requesting the trusted score of the trusted object to the trusted object in the social network. Path reliability inference step; Each of the path reliability scores inferred for each of the plurality of paths from the trusted subject to the trusted object is weighted and summed to calculate a final reliability score of the trusted object for the trusted subject, wherein the respective path reliability scores are calculated. A confidence weighting step in which a weight assigned to is set to a value inversely proportional to a difference from an average of the plurality of path reliability scores; And a reputation inference step of inferring a reputation score of the trust object in the social network by summing up the final reliability scores calculated for nodes having a path connected to the trust object.

According to the apparatus and method for trust and reputation inference in a web-based social network according to the present invention, when the trust subject and the trust object are connected to each other through a plurality of paths, By calculating the final reliability score of the trusted object by weighted sum, it is possible to increase the contribution of values close to the average of the plurality of path reliability scores and to improve the accuracy of the final reliability score. In addition, when inferring the reputation of a trust object, only the final confidence scores for nodes located within a certain range from the trust object are summed to eliminate the confidence scores for nodes that are too far from the trust object, improving accuracy while reducing computational load. Can be.

1 is a block diagram showing the configuration of a preferred embodiment of a reliability and reputation inference apparatus in a web-based social network according to the present invention;
2 illustrates an example of a web-based social network;
3A and 3B are diagrams illustrating one path connecting a trust subject and a trust object, and three nodes used for path reliability inference, respectively;
4 is a diagram illustrating pseudo code for inferring a confidence score between a trust subject and a trust object connected by one path;
5 is a diagram illustrating a plurality of paths to a trusted object;
6 is a diagram illustrating a pseudo code for calculating a final reliability score by summing a plurality of path reliability scores;
7 illustrates a relationship with other nodes in a social network to infer the reputation score of a trusted object,
8 is a diagram illustrating a method of inferring reputation of a trusted object in pseudo code;
9 shows a fuzzy graph for reliability and plate;
10 shows a fuzzy graph based on the results of Table 1;
11 is a flowchart illustrating a preferred embodiment of a method of inferring a reputation and reputation in a web-based social network according to the present invention;
12 is a graph showing inferred reliability and accuracy of a flat plate according to the value of R _A.

Hereinafter, exemplary embodiments of a reliability and reputation inference apparatus and method in a web-based social network according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram showing the configuration of a preferred embodiment of a reliability and reputation inference apparatus in a web-based social network according to the present invention.

Referring to FIG. 1, the reliability and plate inference apparatus according to the present invention includes a path reliability inference unit 110, a reliability adding unit 120, a plate inference unit 130, and a credit determination unit 140. In addition, the reliability and reputation inference apparatus according to the present invention is implemented to evaluate the online transaction creditworthiness of a trust object selected from a plurality of nodes constituting a web-based social network.

The path reliability inference unit 110 calculates a path reliability score of the trust object for the trust entity based on the reliability between intermediate nodes located on the path from the trust entity that requested the trust score of the trust object to the trust object in the social network. Infer.

Web-based social networks are created to represent relationships between people or groups within the online trading system of interest. Each member, or node, in the network represents a user of an online trading system, and a tie represents a relationship between users. The quantitative score given to each tie represents the trust between two users connected by that tie. Reliability and reputation inference apparatus according to the present invention is another node (hereinafter referred to as "trust object") of one of the plurality of nodes constituting such a web-based social network (hereinafter referred to as "trust object") Can be used to infer the reliability and reputation of a trusted object within a social network.

Every node in the network assigns a subjective reliability score to other nodes directly connected to it, and the reliability score given is a ratio between 0 and 10. A confidence score of zero means that no connection exists between the two nodes. If the trust subject and the trust object are not directly connected, the trust and reputation of the trust object can be inferred by searching the path between the trust subject and the trust object to the trust object including intermediate nodes directly connected to each other.

2 illustrates an example of a web-based social network. In FIG. 2, the node denoted by 'P' is directly connected to two nodes in the network and indirectly connected to four other nodes. Nodes that are indirectly connected to node 'P' are eventually connected to node 'P' via one or more intermediate nodes. As can be seen in FIG. 2, there may be a plurality of paths connecting two nodes in a social network. Each path is also given a different confidence score.

The path reliability inference unit 110 infers the path reliability score of the trust object for the trust object for each of the plurality of paths connecting the two nodes, that is, the trust object and the trust object. The reliability score inferred for the specific path connecting the trust subject and the trust object is hereinafter referred to as the 'path reliability score'.

When there are a plurality of paths between the trust subject and the trust object, the combination of nodes constituting each path is different from each other, and thus the inferred path reliability scores are also different. Also, if the trust principal and the trust object are directly connected, the trust score given by the trust principal is the path trust score for the corresponding path, but if the intermediate node is included in the path connecting the trust principal and the trust object, Since there is no information about the path reliability score of the trusted object, it should be inferred.

3A and 3B are diagrams illustrating one path connecting a trust subject and a trust object, and three nodes used for path reliability inference, respectively. Referring to FIG. 3A, three intermediate nodes are included in the path from the trust subject to the trust object. The path reliability score of the trust object is inferred through the process of repeatedly accumulating the reliability of the intermediate node existing on the path toward the trust object with the initial value of the reliability of the intermediate node directly connected to the trust subject.

3B shows three nodes located consecutively on the path from the trust subject to the trust object. In FIG. 3B, a node denoted by U ₁ represents a trust entity, and U ₂ and U ₃ represent nodes directly connected to the trust entity and nodes directly connected to the U ₂ node, respectively.

Based on the trust subject U ₁ , the reliability score of U ₂ given by U ₁ is the reliability score of the direct connection, while the reliability score of U ₃ granted by U ₂ is the reliability score of the indirect connection. This is because U ₁ and U ₃ are not directly connected, but indirectly through U ₂ . In U ₁ U ₂ to the confidence score, and U of the U ₂ by directly connecting _two from U ₃ by a U ₃ confidence score by indirect connection to U ₃ confidence score by direct connection to the U ₃ from U ₁ to Inference can be expressed as Equation 1 below.

Here, T _r is at U ₁ U of ₃ to U by direct connection ₃ confidence score, T _d is at U ₁ U ₂ to the reliability of the U ₂ by direct connection score, T _i is indirectly from U ₂ to U ₃ The reliability score of U ₃ by the connection, and V _max is the maximum possible reliability score. That is, when the reliability is expressed as a rational number between 0 and 10, V _max becomes 10.0.

As can be seen from Equation 1, in inferring the path reliability score of the trust object for the trust subject, the trust score by the direct connection and the trust score by the indirect connection are not treated equally, and are directly given by the trust subject. Reliability scores by direct connection play a more important role in reliability inference. Therefore, to adjust the importance level between T _d and T _i , a weighting scheme is used to add the complement of T _d , ie, (V _max -T _d ), to T _i .

Now that the reliability score of U ₃ is calculated by a direct connection from U ₁ , a trust subject, to U _3, the equation 1 above is the node that follows U ₃ on the path from the trust subject to the trust object, for example U _4. Apply repeatedly for. In this case, in applying Equation 1 to U ₄ , the reliability score of U ₃ by the direct connection from U ₁ to U ₃ is T _d , and the reliability score of U ₄ by the indirect connection from U ₃ to U ₄ is The reliability score of U ₄ by T _i , and the new direct connection from U ₁ to U ₄ is T _r .

The path reliability inference unit 110 repeatedly applies the above process to each intermediate node on the path until the path reliability score of the trust object by the direct connection from the trust subject to the trust object is calculated. 4 shows a pseudo code for inferring a confidence score between a trust subject and a trust object connected by one path.

The reliability adder 120 weights and sums each of the inferred path reliability scores for each of the plurality of paths from the trust subject to the trust object, and calculates a final reliability score of the trust object for the trust subject, respectively. The weight given to the reliability score is set to a value inversely proportional to the difference from the average of the plurality of path reliability scores.

There may be a plurality of paths connecting a trust subject and a trust object in a web-based social network. Therefore, in order to infer the final reliability score of the trust object for the trust subject, it is necessary to combine all the path reliability scores calculated for the plurality of paths. The combined reliability score may be given by the trust principal when the trust subject and the trust object are directly connected, or the path trust inference unit 110 when the trust subject and the trust object are connected by a path including an intermediate node. It may be calculated by.

When calculating the final reliability scores by summing the path reliability scores inferred by the plurality of paths connecting the trust subject and the trust object, the present invention is based on the average of the path reliability scores instead of simply adding up the plurality of path reliability scores. Use one weighted sum. That is, the final reliability score is calculated by assigning a weight to each of the plurality of path reliability scores and giving a higher weight to a value close to the average of the plurality of path reliability scores.

5 is a diagram illustrating a plurality of paths to a trust object. Referring to FIG. 5, there are i paths connecting the trust subject and the trust object, and a path reliability score corresponding to each path is calculated as T ₁ to T _i . The confidence summing unit 120 calculates a final confidence score of the trust object for the trust subject by assigning a weight of W ₁ to W _i to the path reliability scores of T ₁ to T _i , respectively. This is represented by Equation 2 below.

Where T _R is the final confidence score of the trust object for the trust entity, T _i is the path reliability score inferred for the i th path connecting the trust entity and the trust object, and W _i is given to the path reliability score of the i th path The weight, and avg (T _n ), is the average of all path reliability scores. In addition, the sum of all weights can be 1.0.

According to Equation 2, the plurality of path reliability scores are closer to the average value, and are assigned a higher weight when calculating the final reliability score. By using such weighted method, the influence of the extreme reliability score on the final reliability score can be reduced, and the reliability of the calculated reliability score can be increased. 6 is a diagram illustrating a pseudo code for calculating a final reliability score by summing a plurality of path reliability scores.

The reputation inference unit 130 infers the reputation score of the trust object in the social network by summing the final reliability scores calculated for each node having a path connected to the trust object.

Reputation is another important factor besides reliability for successful online transactions. Every member in a social network has its own reputation, which is formed by the gathering of the personal trust of one or more other members. In other words, unlike trust, which is a subjective measure of trust objects within a social network, trust is an objective measure that trust objects have for all other members of the social network. Therefore, in order to infer the reputation score of a trust object in a social network, the final trust scores of trust objects for all other nodes directly or indirectly connected to the trust object may be summed.

7 is a diagram illustrating the relationship with other nodes in a social network to infer the reputation score of a trust object. Referring to FIG. 7, nodes in a social network directly and indirectly connected to the trust object P are indirectly connected to the trust object through a path including a group of nodes G ₁ directly connected to the trust object and one intermediate node. A group of connected nodes G ₂ and other groups of nodes G _i indirectly connected with a trust object via a path including a greater number of intermediate nodes. I is a value that increases as the number of intermediate nodes increases. These nodes all have information about the final confidence score of the trust object, and the reputation of the trust object is deduced by combining these final confidence scores.

At this time, a node far from the trusted object, that is, a node in which many intermediate nodes are included in the path to the trusted object, does not take much weight in inferring the reputation score of the trusted object. Therefore, in order to consider only the final confidence score of nodes within a certain range from the trust object in the reputation score inference, a value for limiting the range of nodes may be set. That is, by setting the range limit value l in advance, the reputation score of the trusted object is deduced by summing only the final reliability scores of the nodes belonging to the group from G ₁ to G _l . This eliminates nodes that are too far from the trusted object and reduces the computational load. 8 is a diagram illustrating a method of inferring a reputation object in a pseudo code.

The credit determination unit 140 determines a credit degree for determining whether the trust subject is online with the trust object based on the final trust score of the trust object and the reputation score of the trust object in the social network.

Reliability and reputation inferred by the same method as described above are combined with each other for reliable online transactions. Subjective characteristics, such as credibility, are more important in the nature of online transactions, and objective characteristics, such as reputation, are more important in some cases. In some cases, subjective and objective characteristics are considered equal. To account for all of these cases, we can use fuzzy logic and construct a computational model that combines reliability and reputation. In this model, two factors of credibility, reliability and reputation, become fuzzy descriptors.

9 shows a fuzzy graph for reliability and plate. (A) of FIG. 9 is a graph showing the reliability divided into four sets of {low}, {medium}, {high}, and {very high}, and (b) shows the reputation of {very low} and {low }, Graphed with five sets of {medium}, {high}, and {very high}. This classification may vary depending on the members and types of online transactions. The values of confidence and reputation for the trust object calculated by the method described above are mapped to a graph as shown in FIG. 9, and the most appropriate value, for example, the highest value is selected.

Table 1 below shows the results of the reliability and reputation values.

Reliability reputation result One lowness Very low Very low 2 lowness lowness lowness 3 lowness middle Slightly lower 4 lowness height middle 5 lowness Very high Slightly higher 6 middle Very low lowness 7 middle lowness Slightly lower 8 middle middle middle 9 middle height Slightly higher 10 middle Very high height 11 height Very low Slightly lower 12 height lowness middle 13 height middle Slightly higher 14 height height height 15 height Very high Very high 16 Very high Very low Slightly lower 17 Very high lowness middle 18 Very high middle Slightly higher 19 Very high height height 20 Very high Very high Very high

Table 1 is based on the values shown in the graph of FIG. 9, indicating that there are 20 cases of 4-scale reliability and 5-scale flat plate. Referring to Table 1, a result value considering both reliability and reputation can be obtained.

10 is a diagram illustrating a fuzzy graph based on the result values of Table 1. FIG. The graph of FIG. 10 is obtained by inferring the quantitative steps of creditworthiness by simultaneously considering reliability and reputation. As also shown in Table 1, credit ratings are classified into seven sets of {very low}, {low}, {slightly low}, {medium}, {slightly high}, {high}, and {very high}. From this fuzzy graph, one can infer the creditworthiness of a particular member of a social network.

11 is a flowchart illustrating a preferred embodiment of a method for inferring reputation and reputation in a web-based social network according to the present invention.

Referring to FIG. 11, the path reliability inference unit 110 trusts based on a reliability between intermediate nodes located on a path from a trust subject that requests a trust score of a trust object to a trust object in a web-based social network. Infer the path reliability score of the trusted object with respect to (S1010). However, when the trust subject and the trust object are directly connected, the trust score of the trust object given by the trust subject may be used as the path reliability score.

The confidence summing unit 120 calculates a final confidence score of the trust object for the trust entity by weighting and summing each of the inferred path reliability scores for each of the plurality of paths from the trust subject to the trust object (S1020). The weight given to each path reliability score is set to a value inversely proportional to the difference from the average of the plurality of path reliability scores.

The reputation inference unit 130 infers the reputation score of the trust object in the social network by summing the final reliability scores calculated for each node having a path connected to the trust object (S1030). At this time, in order to infer the reputation score, only the final reliability scores of nodes connected by a path including intermediate nodes within a predetermined number between the trust object and the trusted object may be used to improve the accuracy of the reputation score and reduce the computational load.

Finally, the credit determination unit 140 determines the creditworthiness for determining whether or not the subject trusts the on-line transaction with the trust object based on the final trust score of the trust object and the reputation score of the trust object in the social network. (S1040). Fuzzy logic can be used to determine creditworthiness.

Experiments were conducted to evaluate the performance of the present invention. Every member of a web-based social network has a confidence score for other directly connected members.

The accuracy of the reliability inferred according to the present invention is evaluated by considering only the relationship between the trust subject and the trust object. Social networks are created to evaluate the reliability inference model proposed by the present invention and are given various characteristics for each member in the social network. Each member corresponding to 500 nodes in the created social network is directly connected with an average of 20 nodes, and the confidence score for this directly connected node is determined by evaluating the similarity of the characteristics of both nodes. The accuracy of the reliability inference model according to the present invention is evaluated by comparing a given confidence score based on the similarity of the inferred reliability score and the characteristic. The experiment was repeated 1000 times.

The accuracy of the present invention was compared with the conventional reliability inference model 'TidalTrust', and the average accuracy of the present invention and 'TidalTrust' is shown in Table 2 below.

Average accuracy Invention 97.2% Tidaltrust 93.4%

Referring to Table 2, the reliability inference model according to the present invention can be seen that the accuracy is improved by 3.8% compared to the 'TidalTrust' model. In the 'TidalTrust' model, the inference accuracy decreases as the length of the path between two nodes, the trust principal and the trust object, increases. On the other hand, the present invention maintains a constant accuracy regardless of the length of the path.

Since the reliability scores obtained by the reliability inference model according to the present invention are also summed when the plate is inferred, it can be seen that the accuracy of the reliability inference model according to the present invention reflects the accuracy of the plate inference model according to the present invention.

The second experiment was performed to evaluate the accuracy of the reliability inference model, reputation inference model, and reliability score summing method according to the present invention. Each node in the social network is given a preset confidence score (hereinafter referred to as a 'standard value'). The standard value is used as a reference for comparing the reliability scores obtained using the reliability inference model according to the present invention. By comparing the inferred reliability score and the standard value, it is possible to evaluate the accuracy of the reliability inference model according to the present invention.

As in the first experiment, each node in the social network gives subjective confidence scores to other directly connected nodes. However, the difference from the first experiment is that the confidence score given is proportional to the standard value S _V and the accuracy R _A of the given score. Inferred reliability and reputation scores are compared to standard values to assess the accuracy of the present invention.

For the experiment, 500 nodes are created in the web-based social network, and the standard value assigned to each node has a normal distribution of 5.0 on average. When a standard value is assigned to all nodes in a social network, a confidence score is assigned to directly connected nodes based on R _A and S _V. For example, if R _A is 100%, the confidence score given to the directly connected node is equal to the S _V value of the node. If R _A decreases by 10%, the difference between the reliability score given to the node and the S _V value of the node increases to 1.0 (± 0.5). The accuracy of inferred confidence and reputation scores is assessed by varying the value of R _A from 0 to 100%.

12 is a graph showing inferred reliability and accuracy of a flat plate according to the value of R _A. 12A is a graph showing the accuracy of the reliability value, and (b) is a graph showing the accuracy of the plate value. Referring to FIG. 12, when the value of R _A is low, the inferred reliability and accuracy of the flat plate are also relatively low. However, as the value of R _A increases, the accuracy also improves. This result means that the value of R _A set by the user greatly affects the accuracy of the present invention. Since the confidence score inferred for one path depends on the selected path, the accuracy graph of FIG. 12 does not increase constantly.

Table 3 below shows the average difference between the standard value S _V and the inferred reliability and reputation scores, ie average accuracy, as R _A increases.

R _A Average accuracy of reliability Average accuracy of the plate 80% 81.142% (0.9429) 79.746% (1.0127) 85% 85.694% (0.7153) 85.164% (0.7418) 90% 90.542% (0.4729) 90.918% (0.4541) 95% 92.832% (0.3584) 93.290% (0.3355) 100% 96.198% (0.1901) 96.410% (0.1795)

Referring to Table 3, since the average error increases as the value of R _A increases, the difference between S _V and the inferred reliability and reputation values decreases. In other words, as the value of R _A increases, the accuracy of the inference of the present invention increases, and approaches the value of R _A. When R _A is 100%, the accuracy of the reliability inference model according to the present invention is about 96%, and the accuracy of the plate inference model is about 97%.

In addition, Table 4 below shows the reliability and accuracy of the plate inference model according to the present invention when R _A is 100% compared with the conventional inference model Golbeck's model.

Reliability reputation Invention 96.198% 96.410% Model of Golbeck 91.286% 92.572%

Referring to Table 4, it can be seen that the reliability and the accuracy of the plate inferred in the case of using the present invention are improved by about 4 to 5% compared to the conventional methods.

The present invention can also be embodied as computer-readable codes on a computer-readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data that can be read by a computer system is stored. Examples of the computer-readable recording medium include a ROM, a RAM, a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like, and may be implemented in the form of a carrier wave (for example, transmission via the Internet) . The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

Although the preferred embodiments of the present invention have been shown and described above, the present invention is not limited to the specific preferred embodiments described above, and the present invention belongs to the present invention without departing from the gist of the present invention as claimed in the claims. Various modifications can be made by those skilled in the art, and such changes are within the scope of the claims.

110-Path Reliability Inference
120-reliability adder
130-reputation inference
140-Credit Decision Unit

Claims (13)

In the reliability and reputation inference device for evaluating the online transaction credit of the selected trust object among a plurality of nodes constituting the web-based social network,
A path for inferring a path reliability score of the trusted object for the trusted object based on the reliability between the intermediate nodes located on the path from the trusted subject that requested the credit score of the trusted object to the trusted object in the social network; Reliability inference unit;
Each of the path reliability scores inferred for each of the plurality of paths from the trusted subject to the trusted object is weighted and summed to calculate a final reliability score of the trusted object for the trusted subject, wherein the respective path reliability scores are calculated. A reliability adder configured to set a weight assigned to the value in inverse proportion to a difference from an average of the plurality of path reliability scores; And
And a reputation inference unit for inferring a reputation score of the trust object in the social network by summing final final confidence scores calculated for nodes having a path connected to the trust object. Inference Device. The method of claim 1,
A credit determination unit configured to determine a credit degree for determining whether the trust subject is online with the trust object based on the final trust score of the trust object and the reputation score of the trust object in the social network. Reliability and flat plate inference device, characterized in that it further comprises. The method of claim 2,
The credit determination unit may apply a fuzzy logic to map the value of the final credit score to any one of a preset number of reliability levels, and map the value of the reputation score to any one of a preset number of reputation levels. Reliability and reputation inference device, characterized in that determining the credit rating. 4. The method according to any one of claims 1 to 3,
The path reliability inference unit calculates a path reliability score of the trust object by repeatedly calculating a reliability score of the trust object sequentially for nodes included along the path from the trust object to the trust object by Equation A below. Reliability and reputation inference device, characterized in that:
Equation A

Here, T _d is a confidence score of the trust subject of the first node of the nodes included in the path, T _i is the reliability of the second node of the second node directly connected to the first node along the path. The score, T _r, is the confidence score for the trust subject of the second node, and V _max is the maximum value of the confidence score. 4. The method according to any one of claims 1 to 3,
The reliability summarizing unit calculates the final reliability score according to Equation B below.
Equation B

Here, T _R is the final reliability score, T _i is a path reliability score inferred for the i-th path among a plurality of paths connecting the trust subject and the trusted object, W _i is the path reliability score of the i-th path The weight given, V _max is the maximum value of the reliability score, and avg (T _n ) is the average of the path reliability scores inferred for the plurality of paths, and the sum of the weights assigned to the plurality of path reliability scores is 1.0 to be. 4. The method according to any one of claims 1 to 3,
The reputation inference unit infers the reputation score by summing up the final reliability scores calculated for each of the nodes connected through the path including the trust object and a predetermined number of intermediate nodes or less. Device. In the reliability and reputation inference method for evaluating the online transaction credit of the selected trust object among a plurality of nodes constituting the web-based social network,
A path for inferring a path reliability score of the trusted object for the trusted object based on the reliability between the intermediate nodes located on the path from the trusted subject that requested the credit score of the trusted object to the trusted object in the social network; Reliability inference step;
Each of the path reliability scores inferred for each of the plurality of paths from the trusted subject to the trusted object is weighted and summed to calculate a final reliability score of the trusted object for the trusted subject, wherein the respective path reliability scores are calculated. A confidence weighting step in which a weight assigned to is set to a value inversely proportional to a difference from an average of the plurality of path reliability scores; And
And a reputation inference step of inferring a reputation score of the trusted object in the social network by summing final final confidence scores calculated for each node having a path connected to the trusted object. Reasoning method. The method of claim 7, wherein
A credibility determination step of determining a credit degree for determining whether the trust subject makes an online transaction with the trust object based on the final trust score of the trust object and the reputation score of the trust object in the social network. Reliability and reputation inference method characterized in that it further comprises. The method of claim 8,
In the credit determination step, fuzzy logic is applied to map the value of the final confidence score to any one of a preset number of confidence levels, and the value of the reputation score is mapped to any one of a preset number of reputation levels. Reliability and reputation inference method characterized in that for determining the credit. The method according to any one of claims 7 to 9,
In the path reliability inference step, path reliability of the trusted object is repeatedly calculated by sequentially calculating the reliability scores for the trusted subjects sequentially for nodes included along the path from the trusted subject to the trusted object by Equation A below. Reliability and reputation inference methods characterized by inferring scores:
Equation A

Here, T _d is a confidence score of the trust subject of the first node of the nodes included in the path, T _i is the reliability of the second node of the second node directly connected to the first node along the path. The score, T _r, is the confidence score for the trust subject of the second node, and V _max is the maximum value of the confidence score. The method according to any one of claims 7 to 9,
In the reliability summing step, the reliability and reputation inference method characterized in that to calculate the final reliability score by the following equation B:
Equation B

Here, T _R is the final reliability score, T _i is a path reliability score inferred for the i-th path among a plurality of paths connecting the trust subject and the trusted object, W _i is the path reliability score of the i-th path The weight given, V _max is the maximum value of the reliability score, and avg (T _n ) is the average of the path reliability scores inferred for the plurality of paths, and the sum of the weights assigned to the plurality of path reliability scores is 1.0 to be. The method according to any one of claims 7 to 9,
In the reputation inference step, the reliability score is inferred by summing up the final reliability scores calculated for each of the nodes connected through the path including the trust object and a predetermined number of intermediate nodes or less. Reputation reasoning method. A computer-readable recording medium having recorded thereon a program for executing the reliability and reputation inference method according to any one of claims 7 to 9.

Images