KR102042040B1 - Method of maximizing reverse influence in social-network with negative influence - Google Patents

Abstract

The present invention relates to a method for maximizing an adverse effect in a social network having a negative influence capable of performing cost-benefit-analysis while maximizing a marketing influence which comprises: a parameter determining step; a seeding cost calculation step of calculating seeding costs for all neighboring nodes of a target node; and a total seeding cost calculation step of calculating the total seeding costs.

Description

Method of maximizing reverse influence in social-network with negative influence

The present invention relates to a method for maximizing adverse effects in a social network having a negative influence. More particularly, the present invention relates to a seeding cost estimation method in a social network having a negative influence, and a computing device and a system for performing the same.

Influence Maximization (IM) is a potential model for viral marketing in social network research for many years. IM aims to find a small set of seed users within the network that can maximize impact and benefit.

On the other hand, seeding cost of viral marketing is estimated by Reverse Influence Maximization (RIM), in which the influence is transmitted in the reverse direction. Seeding cost is the minimum number of objects to be activated required to activate all users. IM and RIM are very important in research and business because they can provide cost-benefit analysis.

Kempe et al. As a pioneer in this research, he proposed the Linear Threshold (LT) and Independent Cascade (IC) models. Leskovec et al. Proposed a widely used Cost-Effective Lazy Forward (CELF) model for large-scale infection detection, which outperformed many existing greedy models. Gardner et al. Used the Susceptible-Exposed-Infected-Recovered (SEIR) model for viral marketing.

However, none of the above-mentioned methods has a problem that the seeding cost of viral marketing has not been found. Meanwhile, Talukder et al. First introduced and extended the RIM model. However, the existing Reverse Influence Maximization (RIM) model has a problem that RIM has not solved the stopping criteria, basic network components, and insufficient impact problems. In addition, the IM and RIM models mentioned above have a problem that they do not consider negative effects in social networks.

Summary of the Invention The present invention is to solve the above-mentioned problems, and to estimate the optimization seeding cost and opportunity cost of viral marketing on social networks.

In addition, the technical problem to be achieved by the present invention is to provide a method for determining the cost and online advertising costs for maximizing the profit of viral marketing in social networks.

In a seeding cost estimation method in a social network according to an aspect of the present invention, the method is performed by an estimation apparatus. The method comprises the steps of: determining, for a node consisting of a plurality of hops on the social network, a contact rate (a), an infection rate (b) and a recovery rate (c); A seeding cost calculating step of calculating seeding cost γ (u) for all neighboring nodes u of the target node v; And a total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈ S} based on the optimized seeding cost for activating the target node v of the seeding cost γ (u). _' Γ (v) | The cost-benefit-analysis can be performed while maximizing the marketing impact on the social network through cost estimation and cost optimization techniques.

In one embodiment, the seeding cost calculation step may include: an initialization step of initializing R = 0 and I = 1, assuming that no individual has recovered from an initial state, and then the neighboring node u is activated or infected; And counting the number of nodes in the susceptible, exposed (E), infected (I) and infected (R) recovery groups up to the m-th hop.

로 계산될 수 있다. 여기서, It는 m번째 홉의 전염 검사 시간에서의 감염 및 회복에 따른 차이를 고려한 감염 노드 수에 해당하고, 상기 차이를 고려한 감염 노드 수에 기반하여 상기 시딩 비용 γ(u) 이 업데이트될 수 있다.In one embodiment, the number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R) is

It can be calculated as Here, it corresponds to the number of infected nodes considering the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) can be updated based on the number of infected nodes considering the difference. .

에 의해 상기 총 시딩 비용이 최소값을 갖도록 결정될 수 있다.In one embodiment, at the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈ S '} Γ (v) |

The total seeding cost can be determined to have a minimum value.

In one embodiment, the step of calculating the number of nodes, calculating the number of infected nodes It considering the difference according to the infection and recovery until the infection rate (b) is less than the recovery rate (c), the seeding cost γ (u ) Can be updated.

In an embodiment, the most influential user in the social network corresponding to the target node v may be selected, and a promotion request message for the corresponding product or service may be transmitted to the user terminal of the user.

In an embodiment, in the calculating the total seeding cost, selecting two or more of the neighboring nodes such that the total seeding cost among the first neighboring nodes directly connectable with the target node v has a minimum value, and the target node The method may further include a profit prediction step of performing profit prediction based on the second neighboring nodes connectable to (v).

A computing device is provided for performing seeding cost estimation in a social network in accordance with another aspect of the present invention. The computing device includes an interface for receiving information about a directed graph G (V, E) having an interaction set E relating to the interaction of the individual set V with the individual set V; And for a node consisting of a plurality of hops on the social network, determine a contact rate (a), an infection rate (b) and a recovery rate (c), and seeding for all neighboring nodes u of the target node v. Compute the cost γ (u) and based on the optimized marginal seeding cost for activating the target node v of the seeding cost γ (u), the total seeding cost γ (S ') = | Γ (S' ) | = | ∪ _{v∈ S '} Γ (v) |

In one embodiment, the processor initializes R = 0, I = 1, and is vulnerable to the mth hop, assuming that no individual has recovered from its initial state, and then the neighbor node u has been activated or infected. The number of nodes in the (S: susceptible), exposed (E: exposed), infected (I: affected), and recovery (R) recovery groups can be calculated.

로 계산될 수 있다. 여기서, It는 m번째 홉의 전염 검사 시간에서의 감염 및 회복에 따른 차이를 고려한 감염 노드 수에 해당하고, 상기 차이를 고려한 감염 노드 수에 기반하여 상기 시딩 비용 γ(u) 이 업데이트될 수 있다.In one embodiment, the number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R) is

It can be calculated as Here, it corresponds to the number of infected nodes considering the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) can be updated based on the number of infected nodes considering the difference. .

에 의해 상기 총 시딩 비용이 최소값을 갖도록 결정할 수 있다.In one embodiment, the processor is configured such that at the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈S '} Γ (v) |

It can be determined that the total seeding cost has a minimum value.

In one embodiment, the processor calculates the number of infected nodes It considering the difference according to the infection and recovery until the infection rate b is less than the recovery rate c, and updates the seeding cost γ (u). can do.

In one embodiment, the processor selects the most influential user from the social network corresponding to the target node v, and sends the promotion request message for the product or service to the user terminal of the user. You can control the interface.

In one embodiment, the processor selects two or more of the neighboring nodes such that the total seeding cost among the first neighboring nodes directly connectable with the target node v has a minimum value, and selects the target node v. Profit prediction may be performed based on connectable second neighbor nodes.

According to another aspect of the present invention, a computer-readable storage medium having recorded thereon program code for performing seeding cost estimation in a social network is provided. The program code causes the computer to determine a contact rate (a), an infection rate (b) and a recovery rate (c) for a node consisting of a plurality of hops on the social network, and cause the computer to determine a target node ( calculate seeding cost γ (u) for all neighboring nodes u of v), and cause the computer to optimize the marginal seeding cost for activating the target node v of the seeding cost γ (u). Based on the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈S '} Γ (v) | Calculate

In one embodiment, the program code initializes R = 0, I = 1, assuming that the computer has not recovered any individuals from the initial state, and then that the neighbor node u has been activated or infected, Allow the computer to count the number of nodes in the susceptible, exposed (E), infected (I) and infected (R) recovery groups up to the mth hop.

로 계산하도록 할 수 있다. 여기서, It는 m번째 홉의 전염 검사 시간에서의 감염 및 회복에 따른 차이를 고려한 감염 노드 수에 해당하고, 상기 차이를 고려한 감염 노드 수에 기반하여 상기 시딩 비용 γ(u) 이 업데이트될 수 있다.In one embodiment, the program code may determine the number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R).

Can be calculated as Here, it corresponds to the number of infected nodes considering the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) can be updated based on the number of infected nodes considering the difference. .

에 의해 상기 총 시딩 비용이 최소값을 갖도록 결정할 수 있다.In one embodiment, the program code is at the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈S '} Γ (v) |

It is possible to determine that the total seeding cost has a minimum value.

A system for performing seeding cost estimation in a social network in accordance with another aspect of the present invention is provided. The system includes a user terminal in which a social network application is installed; And a server for receiving information on the directed graph G (V, E) having the interaction set E relating to the interaction between the individual set V of nodes corresponding to the user terminal and the individual set V. FIG. Here, the server determines a contact rate (a), an infection rate (b) and a recovery rate (c) with respect to a node consisting of a plurality of hops on the social network, and determines all neighboring nodes of the target node v ( calculate seeding cost γ (u) for u) and, based on the optimized marginal seeding cost for activating the target node v of the seeding cost γ (u), total seeding cost γ (S ') = | Γ (S ') | = | ∪ _{v∈ S'} Γ (v) | Calculate

In one embodiment, the server initializes R = 0, I = 1 and is vulnerable to the mth hop, assuming that no individual has recovered from its initial state and then the neighbor node u has been activated or infected. The number of nodes in the (S: susceptible), exposed (E: exposed), infected (I: affected), and recovery (R) recovery groups can be calculated.

The expected effect of the present invention is that the cost estimation and cost optimization techniques proposed by the present invention enable all organizations to perform cost-benefit analysis while maximizing their impact.

 In addition, the expected effect according to the present invention is that the IM provides a profit estimate and the EN-RIM provides a cost calculation so that both can be combined for cost-benefit analysis.

In addition, the expected effect according to the present invention has the advantage of applying the model according to the present invention in the budgeted advertisement coverage problem to promote the product / service at a minimum cost by selecting the most influential user in the social network.

BRIEF DESCRIPTION OF THE DRAWINGS In order to better understand the drawings cited in the detailed description of the invention, a brief description of each drawing is provided.
1 shows a SEIF model and a flowchart according to the present invention.
2 illustrates a system model including nodes corresponding to a plurality of users in relation to a seeding cost estimation method in a social network according to the present invention.
3 is a flowchart of a seeding cost estimation method in a social network having a negative effect according to the present invention.
4 illustrates a connection state between a target node and a neighbor node connected to the target node associated with activation of the target node considering the optimal cost according to the present invention.
5 shows a detailed configuration of a system including a computing device for performing seeding cost estimation in a social network according to the present invention.

 The present invention may be variously modified and have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail with reference to the accompanying drawings. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

In describing the present invention, when it is determined that the detailed description of the related known technology may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, numerals (eg, first, second, etc.) used in the description process of the present specification are merely identification symbols for distinguishing one component from another component.

In addition, in the present specification, when one component is referred to as "connected" or "connected" with another component, the one component may be directly connected or directly connected to the other component, but in particular It is to be understood that, unless there is an opposite substrate, it may be connected or connected via another component in the middle.

The suffixes "module" and "unit" for components used in the following description are given or used in consideration of ease of specification, and do not have distinct meanings or roles from each other. In addition, in order to clearly demonstrate the present invention, parts irrelevant to the description are omitted in the drawings. Like reference numerals denote like elements throughout the specification.

Hereinafter, with reference to the accompanying drawings will be described in detail for the practice of the present invention.

As described above, the existing Reverse Influence Maximization (RIM) model has a problem in that it does not properly solve the stopping criteria, basic network components, and insufficient impact problems of RIM. In addition, there is a problem that Influence Maximization (IM) and RIM models do not consider negative effects in social networks.

Therefore, the present invention proposes an epidemic-based RIM model with negative influencing (EN-RIM) that has a negative effect in order to maximize the profit in social networks. In the present invention, the SEIR model is modified to suit the seeding cost estimation. Epidemiological models are applicable to viral marketing of social networks based on how the disease infects individuals. The SEIR model can be used to divide populations (individuals, individuals) into different groups. In this regard, Figure 1 shows a SEIF model according to the invention and a flowchart accordingly.

Susceptible (S): This group is not yet infected but is vulnerable to disease. Population for maximum profit.

2. Exposed (E): This population contains a subset of group S that has been exposed to the disease but not yet infected. They may be considered individuals exposed to advertising or personal influence.

3. Infected (I): This group includes individuals with disease and can become infected when susceptible candidates are exposed to this group. In a profit-maximizing application, an individual's infection means that the product has been adopted and is incentive for cost optimization.

4. Recovered (R): This group consists of individuals who have had a disease but have recovered. In a profit-maximizing application, these users decided to adopt the product, but eventually did not adopt it because of the negative impact.

Meanwhile, in the present invention, the dynamic SEIR model is modified to calculate seeding costs (opportunity costs) of viral marketing and target marketing in social networks using the RIM model.

Meanwhile, in relation to a seeding cost estimation method in a social network according to the present invention, a system model including nodes corresponding to a plurality of users will be described as follows. In this regard, FIG. 2 illustrates a system model including nodes corresponding to a plurality of users in relation to a seeding cost estimation method in a social network according to the present invention.

In this regard, the epidemic model was designed using the RIM (EN-RIM) in consideration of the negative impact in order to minimize the seeding cost. The EN-RIM model is applied to viral marketing in social networks represented by directed graph G (V, E) with individual set V and their social interaction set E. We introduce a non-neighbor set n (v) and a neighbor set n- ¹ (v) for each node v.

Meanwhile, in relation to the EN-RIM model according to the present invention, a dynamic SEIR model may be applied at each hop of neighbors of the target node v. The main object of the present invention is to maximize profits by minimizing costs. In this case, if an individual is infected, it means that a cost unit is generated while providing a constant cost incentive in exchange for affecting the seed user. Recovery of an individual meant that he changed his decision to take into account the negative effects. For example, if the price is higher than he expected, or if you see bad reviews from customers who have already used the product.

Recovered individuals reject incentives and are not deceived by viral marketing. The SEIR model was modified and applied to the EN-RIM model. The EN-RIM model first finds the seeding cost γ (u) of all neighbors of v and calculates the optimized marginal seeding cost Γ (v).

For nodes with u∈n ^-1 (v), Ni = | n ^-1 (u) | Assume that no individual has yet recovered in the initial state. That is, R = 0.

Then assume that node u is activated or infected. That is, I = 1.

When calculating for all hops of u's neighbors, the contact rate is a = 1 because all neighbors are assumed to be infected.

Therefore, the infected and potentially exposed individuals are given by Equations 1 and 2 below.

Where t = 1,2,... , m represents the test time of transmission at the m-th hop. At this time, each hop is calculated and the infection test time t = 2, 3,... , for Equation 3 to 5 are derived.

The infected individual at each hop is then calculated to obtain the seed cost set Γ (u), u∈n ^-1 (v) as follows:

Next, as shown in FIG. 4, the target (target) entity v is infected with an infectious disease threshold value θ = 0.5. That is, since at least half of the nodes are selected from n ⁻¹ (v) to activate v, the marginal seeding cost γ (v) = | Γ (v) |

Therefore, γ (S ') is expressed by Equation 8 below

 All such processes are represented by the algorithm of Table 1 and FIG. 3 below.

That is, FIG. 3 shows a flowchart of a seeding cost estimation method in a social network having a negative effect according to the present invention. Referring to FIG. 3, a graph information receiving step S110, a parameter determining step S120, an initialization step S130, a node number calculation step S140, and a total seeding cost calculation step S150 are included. Meanwhile, the initializing step S130 and the node number calculating step S140 may be combined to be referred to as the seeding cost calculation step.

In the step of receiving the graph information (S110), the information on the directed graph G (V, E) having the interaction set E regarding the interaction between the individual set V of nodes and the individual set V is received. On the other hand, in the parameter determination step (S120), for the node consisting of a plurality of hops (hop) on the social network, the contact rate (a), infection rate (b) and recovery rate (c) is determined. At this time, the contact rate (a), the infection rate (b) and the recovery rate (c) may be determined for the node consisting of a plurality of hops based on the information on the directed graphs G (V, E).

In addition, in the initializing step (S130), it is assumed that no individual has recovered from the initial state, and then the neighbor node u is activated or infected, and then initializes to R = 0 and I = 1. Meanwhile, in the node count calculation step (S140), the number of nodes in the susceptible, exposed (E), infected (I: infected), and recovery (R) groups is calculated until the m th hop. Accordingly, in the seeding cost calculation step, the seeding cost γ (u) for all neighboring nodes u of the target node v may be calculated.

On the other hand, the number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R) can be calculated by the following equation (9).

Here, it corresponds to the number of infected nodes considering the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) can be updated based on the number of infected nodes considering the difference. .

에 의해 총 시딩 비용이 최소값을 갖도록 결정될 수 있다.Meanwhile, referring to Equations 7 and 8, in the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈ S '} Γ (v) |

The total seeding cost can be determined to have a minimum value.

On the other hand, in the node count calculation step (S140), until the infection rate (b) is smaller than the recovery rate (c), the number of infected nodes It is considered in consideration of the difference according to the infection and recovery, and the seeding cost γ (u) Can be updated. In this regard, a stop criterion related to the node number calculation step S140 repeatedly performed in FIG. 3 and a method of learning the related parameters will be described below.

Learning rate

 All potentially infected individuals are exposed to infected individuals. A = 1. We use the heuristic technique to learn infection rate b and recovery rate c. The number of heads is counted according to the head / tail probability model by using the probability obtained from the tri-valency model for each node in the potentially suspicious set. This process is calculated for a random number of neighbors on different hops of v and the expected value is obtained as infection rate b. The recovery rate c is also calculated by the same method.

Suspension criteria

The counting of the hops continues until there are no neighboring nodes of the infected node, i.e. no candidate is likely to be infected. However, you can limit the value of t using b <c, which means that the disease disappears after some finite time. In addition, effect of exponential decay function, p _t = p ₀ e ^- can be applied that decreases with the distance with a ^λt. Here, lambda = 1.44 can be applied according to the optimum parameter.

EN-RIM algorithm

O(k²td)) 와 같으며 여기서 d는 d = G 의 최대 차수를 의미한다. 훈련 단계에서 k의 값을 감소시킴으로써 수행 시간을 향상시킬 수 있지만 수행 시간과 훈련 품질을 트레이드 오프하는 결과를 초래할 수 있다.The EN-RIM algorithm adjusts the values of a and b in the fourth line of Table 1 as described in Algorithm 1. Then we find the optimal marginal seeding cost set Γ (v) in lines 5-23 and finally calculate the seeding cost γ (S ') in line 25. Since the training phase has a complexity of O (ktd) and the marginal cost calculation requires O (k) time, the complexity of the overall algorithm is O (k (ktd + k)).

O (k ² td)) where d is the maximum order of d = G. Reducing the value of k in the training phase may improve performance, but may result in tradeoffs in performance and training quality.

Meanwhile, according to an embodiment of the present invention, referring to FIG. 2, a user who is most influential in the social network corresponding to the target node v is selected, and the user terminal of the user for the corresponding product or service. The promotion request message may be sent. In this regard, FIG. 4 shows a connection state between a target node and a neighboring node connected to the target node in relation to activation of the target node considering the optimal cost according to the present invention. Referring to FIG. 4, the target node may be activated such that the total seeding cost among the first neighboring nodes directly connected to the target node v has a minimum value.

Meanwhile, referring to FIGS. 2 and 4, in the calculating of the total seeding cost, the total seeding cost among the first neighboring nodes directly connectable with the target node v is determined to be the minimum value. Two or more of the neighbor nodes may be selected to have. In addition, a profit prediction step of performing profit estimation may be further performed based on second neighboring nodes connectable to the target node v. FIG.

Meanwhile, referring to FIGS. 2 and 3, a detailed method of estimating costs using a system model related to a cost estimating method according to the present invention is as follows.

1 가 되고 첫번째 홉 계산이 끝날 때 I₂=4-1=3 가 된다. 그리고 세번째 홉의 경우, t=3 일 때, N₃=10, I=8, R

2 가 되고 결과적으로 I₃=6 가 된다. 따라서 이제 γ(u₂)=9 임을 알 수 있다. 모든 u∈n^-1 (v)에 대해 같은 방법으로 γ(u) 을 계산한다.As mentioned above, let v be a target node. Seeding cost γ (v) = | Γ (v) | To calculate, first calculate Γ (u) for all u∈n ^-1 (v). Assume that the contact rate a = 1, the infection rate b = 0.8, and the recovery rate c = 0.2. For the second hop, when t = 2, N ₂ = 5, I = 5 * 0.8 = 4, R = 4 * 0.2

1 and I ₂ = 4-1 = 3 at the end of the first hop calculation. And for the third hop, when t = 3, N ₃ = 10, I = 8, R

2 and consequently I ₃ = 6. Thus, we can see that γ (u ₂ ) = 9. Γ (u) is calculated in the same way for all u∈n ⁻¹ (v).

In the last step, when t = 1, the spreading threshold θ = 0.5 is applied (where two nodes of {u ₁ , u ₂ , u ₃ } must be selected). The majority of must be chosen so that γ (v) = | Γ (v) | is chosen to have the minimum value.

Meanwhile, a system including a computing device that performs seeding cost estimation in a social network according to another aspect of the present invention will be described below. In this regard, FIG. 5 shows a detailed configuration of a system including a computing device for performing seeding cost estimation in a social network according to the present invention.

The system includes a computing device 200 that performs seeding cost estimation in a social network according to a user terminal (or node) 100 corresponding to a plurality of nodes and a connection state thereto. Here, the computing device 200 that performs seeding cost estimation in a social network may be referred to as a computer 200 or a server 200. Meanwhile, the computing device 200 may be configured of one destination node and neighbor nodes connected thereto as shown in FIG. 2, or may be configured of a plurality of destination nodes and neighbor nodes respectively connected to them as illustrated in FIG. 5. Meanwhile, in FIG. 5, although the destination node corresponds to an end node for convenience, the destination node may have a node connected to the destination node again as shown in FIG. 2.

Meanwhile, the user terminal 100 is a terminal on which a social network application is installed, and the directed graph G (V, E) as shown in FIG. 2 may be generated by dividing them according to whether the user is a member of the social network application or a connection state. Accordingly, the computing device 200 may estimate the seeding cost according to the (negative) influence on the user terminal 100 corresponding to the plurality of nodes.

The computing device 200 may include an interface 210 and a processor (or controller) 220. The interface 210 is configured to receive information about the directed graph G (V, E) having an interaction set E relating to the interaction of the individual set V with the individual set V.

Accordingly, the processor 220 determines the contact rate (a), the infection rate (b) and the recovery rate (c) for the node consisting of a plurality of hops on the social network, and all neighbors of the target node v. It is configurable to calculate the seeding cost γ (u) for node u. Further, the processor 220 based on the optimal seeding cost for activating the target node v of the seeding cost γ (u), the total seeding cost γ (S ') = | Γ (S') | = | 구성 _{v 을 S '} Γ (v) |

Meanwhile, the processor 220 may initialize R = 0 and I = 1 after assuming that no individual is recovered from the initial state and the neighbor node u is activated or infected. In addition, the processor 220 may calculate the number of nodes in the susceptible, exposed (E), infected (I) and infected (R) recovery groups up to the mth hop.

At this time, the number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R) can be calculated using the above equation (9). Here, it corresponds to the number of infected nodes considering the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) can be updated based on the number of infected nodes considering the difference. .

에 의해 상기 총 시딩 비용이 최소값을 갖도록 결정할 수 있다.On the other hand, the processor 220 at the total seeding cost γ (S ') = | Γ (S') | = | ∪ _v ∪ _{S '} Γ (v) |

It is possible to determine that the total seeding cost has a minimum value.

Meanwhile, the processor 220 may calculate the number of infected nodes It considering the difference according to the infection and recovery until the infection rate b is smaller than the recovery rate c, and update the seeding cost γ (u). have.

On the other hand, the processor 220 selects the most influential user in the social network corresponding to the target node (v), and transmits the promotion request message for the corresponding product or service to the user terminal of the user (interface) 210 can be controlled.

Meanwhile, the processor 220 may select two or more of the neighboring nodes such that the total seeding cost among the first neighboring nodes directly connectable with the target node v has a minimum value. In addition, the processor 220 may perform profit prediction based on second neighboring nodes connectable to the target node v. FIG.

Meanwhile, in the system for performing seeding cost estimation in the social network according to another aspect of the present invention, the above-described server 200 performs the following operations. That is, the server 200 determines a contact rate (a), an infection rate (b) and a recovery rate (c) for a node composed of a plurality of hops on the social network, and all neighbor nodes of the target node v. calculate the seeding cost γ (u) for (u). Further, the server 200 based on the optimized seeding cost for activating the target node v of the seeding cost γ (u), total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈ S '} Γ (v) |

Meanwhile, the server 200 may initialize R = 0 and I = 1 after assuming that no individual is recovered from the initial state and the neighbor node u is activated or infected. In addition, the server 200 may calculate the number of nodes in the susceptible, exposed (E), infected (I) and infected (R) recovery groups up to the mth hop.

Meanwhile, in addition to the above-described operations, the server 200 may perform operations performed by the computing device 200 and operations performed by the estimation method.

Meanwhile, a computer-readable storage medium having recorded therein program code for performing seeding cost estimation in a social network according to another aspect of the present invention will be described below.

In this regard, the program code causes the computer 200 to determine a contact rate (a), an infection rate (b) and a recovery rate (c) for a node consisting of a plurality of hops on the social network, The computer causes the seeding cost γ (u) for all neighboring nodes u of the target node v to be calculated. In this case, the computer 200 includes a plurality of computers or processors for distributed computing in addition to the same computer.

Further, the program code is based on an optimized threshold seeding cost for the computer 200 to activate the target node v of the seeding cost γ (u), so that the total seeding cost γ (S ') = It is possible to calculate | Γ (S ') | = | ∪ _v∈S' Γ (v) |.

On the other hand, the program code may be initialized to R = 0, I = 1, assuming that the computer 200 has not recovered any individual in the initial state, and then the neighbor node (u) is activated or infected. have. In addition, the program code causes the computer 200 to determine the number of nodes in the susceptible, exposed (E), infected (I) and infected (R) recovery groups up to the mth hop. Can be calculated.

In the meantime, the program code causes the computer 200 to calculate the number of nodes in the vulnerability (S), the exposure (E), the infection (I), and the recovery group (R) by using Equation 9 described above. Can be calculated Here, it corresponds to the number of infected nodes considering the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) can be updated based on the number of infected nodes considering the difference. .

에 의해 상기 총 시딩 비용이 최소값을 갖도록 결정할 수 있다.On the other hand, the program code causes the computer 200 to generate the total seeding cost γ (S ') = | Γ (S') | = | ∪ _v∈S'Γ (v) |

It is possible to determine that the total seeding cost has a minimum value.

In addition to the above-described operations, the program code may cause the computer 200 to perform operations performed by the computing device 200 and operations performed by the estimation method.

In the above, the method of estimating seeding cost and a computing device, a computer-readable medium and a system for performing the same in a social network having a negative influence according to the present invention have been described.

On the other hand, in connection with the field of application and application of the present invention, it can be used for the cost estimation for maximizing profits of viral marketing in social networks and for determining the cost of online advertising.

On the other hand, the expected effect of the present invention is that the cost estimation and cost optimization techniques proposed by the present invention allow all organizations to perform cost-benefit analysis while maximizing the impact.

In addition, the expected effect according to the present invention is that the IM provides a profit estimate and the EN-RIM provides a cost calculation so that both can be combined for cost-benefit analysis.

In addition, the expected effect according to the present invention has the advantage of applying the model according to the present invention in the budgeted advertisement coverage problem to promote the product / service at a minimum cost by selecting the most influential user in the social network.

On the other hand, in relation to the prospect of commercialization of the present invention, the RIM model for estimating seeding cost (opportunity cost) of viral marketing proposed by the present invention may be used to maximize influence, maximize profit, and set advertising budget of all products / services in social networks. It can be applied in various fields. The IM model can be applied to forecast potential profits, and the EN-RIM model can be used to predict the cost of making a profit. The model according to the invention is made primarily for practical applications but is well worth the research. Our model has the potential to be part of a maximizing impact study in social network research.

Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments.

The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of the present invention.

100: user terminal
200: computing devices, computers, servers
210: interface
220: processor, controller

Claims (20)

A method for estimating seeding cost in a social network, the method being performed by an estimating apparatus, the method comprising:
A parameter determining step of determining a contact rate (a), an infection rate (b) and a recovery rate (c) for a node consisting of a plurality of hops on the social network;
A seeding cost calculating step of calculating seeding cost γ (u) for all neighboring nodes u of the target node v; And
Based on the optimized seeding cost for activating the target node v of the seeding cost γ (u), the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈ S '} Γ (v) | Seeding cost estimation method comprising the step of calculating the total seeding cost. According to claim 1,
The seeding cost calculation step,
An initialization step of initializing R = 0 and I = 1 assuming that no individual has recovered from the initial state and then the neighboring node u is activated or infected; And
Seeding cost estimation method, comprising counting the number of nodes in the susceptible, exposed (E), infected (I) and infected (R) recovery groups up to the m-th hop . The method of claim 2,
The number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R)

Is calculated as
Here, it corresponds to the number of infected nodes taking into account the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) is updated based on the number of infected nodes considering the difference Seeding cost estimation method. According to claim 1,
_Wherein the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈S '} Γ (v) |

Seeding cost, characterized in that the total seeding cost is determined to have a minimum value. The method of claim 3, wherein
The node counting step,
Seeding cost estimation Way. According to claim 1,
And selecting a user who is most influential in the social network corresponding to the target node (v), and transmitting a promotion request message for the corresponding product or service to the user terminal of the user. According to claim 1,
In the calculating the total seeding cost,
Selecting two or more of the neighboring nodes such that the total seeding cost among the first neighboring nodes directly connectable with the target node v has a minimum value,
And a profit prediction step of performing profit prediction based on second neighboring nodes connectable to the target node (v). A computing device that performs seeding cost estimation in a social network, the computing device comprising:
An interface for receiving information about the directed graph G (V, E) having an interaction set E relating to the interaction of the individual set V with the individual set V; And
For a node consisting of a plurality of hops on the social network, determine the contact rate (a), infection rate (b) and recovery rate (c),
Calculate seeding cost γ (u) for all neighboring nodes u of the target node v,
Based on the optimized seeding cost for activating the target node v of the seeding cost γ (u), the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈ S '} And a processor that calculates Γ (v) |. The method of claim 8,
The processor,
Assuming that no individual has recovered from the initial state and that the neighbor node u has been activated or infected, then initializes to R = 0, I = 1,
A computing device that counts the number of nodes in a susceptible, exposed (E), infected (I), and infected (R) recovery group up to the mth hop. The method of claim 9,
The number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R)

Is calculated as
Here, it corresponds to the number of infected nodes taking into account the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) is updated based on the number of infected nodes considering the difference Computing device. The method of claim 8,
The processor,
_Wherein the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈S '} Γ (v) |

And determine that the total seeding cost has a minimum value. The method of claim 10,
The processor,
Computing device, characterized in that to calculate the number of infected nodes taking into account the difference between the infection and recovery until the infection rate (b) is less than the recovery rate (c), and to update the seeding cost γ (u). The method of claim 8,
The processor,
Selecting the most influential user in the social network corresponding to the target node v, and controlling the interface to transmit a promotion request message for the corresponding product or service to the user terminal of the user, Computing device. The method of claim 8,
The processor,
Selecting two or more of the neighboring nodes such that the total seeding cost among the first neighboring nodes directly connectable with the target node v has a minimum value,
And perform profit prediction based on second neighboring nodes connectable to the target node (v). A computer-readable storage medium having recorded thereon program code for performing seeding cost estimation in a social network, comprising:
The program code is,
Allow a computer to determine a contact rate (a), infection rate (b) and recovery rate (c) for a node consisting of a plurality of hops on the social network,
Let the computer calculate seeding cost γ (u) for all neighboring nodes u of target node v,
The total seeding cost γ (S ') = | Γ (S') | = | ∪ based on the optimized seeding cost for activating the target node v of the seeding cost γ (u). _{v∈S '} Γ (v) | A computer-readable storage medium to calculate a data rate. The method of claim 15,
The program code is,
Assuming that the computer has not recovered any individuals from its initial state, and then that the neighbor node u has been activated or infected, initializes to R = 0, I = 1,
A computer-readable storage medium, which causes the computer to count the number of nodes in the susceptible, exposed (E), infected (I) and infected (R) recovery groups up to the mth hop. . The method of claim 16,
The program code is,
The number of nodes in the vulnerability (S), the exposure (E), the infection (I) and the recovery group (R)

And calculate
Here, it corresponds to the number of infected nodes taking into account the difference according to the infection and recovery at the infection test time of the m-th hop, and the seeding cost γ (u) is updated based on the number of infected nodes considering the difference And a computer-readable storage medium. The method of claim 15,
The program code is,
The total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈S '} Γ (v) | in,

Determining that the total seeding cost has a minimum value. A system for performing seeding cost estimation in a social network,
A user terminal in which a social network application is installed; And
A server receiving information on a directed graph G (V, E) having an interaction set E relating to the interaction between the individual set V of nodes corresponding to the user terminal and the individual set V,
The server,
For a node consisting of a plurality of hops on the social network, determine the contact rate (a), infection rate (b) and recovery rate (c),
Calculate seeding cost γ (u) for all neighboring nodes u of the target node v,
Based on the optimized seeding cost for activating the target node v of the seeding cost γ (u), the total seeding cost γ (S ') = | Γ (S') | = | ∪ _{v∈ S '} Seeding cost estimation system for calculating Γ (v) | The method of claim 19,
The server,
Assuming that no individual has recovered from the initial state and that the neighbor node u has been activated or infected, then initializes to R = 0, I = 1,
Seeding cost estimation system that counts the number of nodes in a susceptible, exposed (E), infected (I) and infected (R) recovery group up to the mth hop.

Images