Journal of Physics: Conference Series

PAPER • OPEN ACCESS You may also like

- Nanotip analysis for dielectrophoretic
Mathematical model of viral marketing concentration of nanosized viral particles
Woon-Hong Yeo, Hyun-Boo Lee, Jong-
Hoon Kim et al.
To cite this article: I Kh Utakaeva 2019 J. Phys.: Conf. Ser. 1353 012122
- Phase transitions in virology
Ricard Solé, Josep Sardanyés and
Santiago F Elena

- Viral nanomechanics with a virtual atomic

View the article online for updates and enhancements. force microscope
María Aznar, Sergi Roca-Bonet and David
Reguera

This content was downloaded from IP address 117.205.202.140 on 14/12/2022 at 16:06

HIRM-2019 IOP Publishing
Journal of Physics: Conference Series 1353 (2019) 012122 doi:10.1088/1742-6596/1353/1/012122

Mathematical model of viral marketing

I Kh Utakaeva
Financial University under the Government of the Russian Federation, 38
Scherbakovskaya, Moscow, Russia

E-mail: [email protected]

Abstract. The article presents the results of studies of mathematical aspects of the viral
marketing. A mathematical model of the spread of a marketing virus as a prefractal graph was
developed. An algorithm for the dissemination of viral information was developed.

1. Introduction
In epidemiology, an epidemic is defined as the spread of an infectious disease in time and space. The
term is intepreted as a process of spreading anything among people (e.g., epidemy of tourism). Viral
marketing is an impact on the audience, promotion of goods and services at the expense and by the
means of the target audience consciously or unconsciously participating in the distribution of
information about a product.
Open advertising is perceived as negative, intrusive and useless information from which people try
to protect themselves. The nature of viral marketing is that original information, the idea as a ready-
made marketing message is virally transmitted as information recommended for viewing. Thus,
information is associated with advantages rather than an imposed advertising appeal. One of the most
effective ways to disseminate information on the Internet is social networks. Social networks are a
powerful tool for influencing public opinion. They influence business, politics and society. Thanks to
the free interaction of a large number of participants, dissemination of information in social networks
has characteristics similar to those of the epidemy. However, classical models of the spread of epidemics
do not take into account peculiarities of viral marketing and social networks. The article analyzes a
mathematical model describing the digital marketing using an epidemiological approach that takes into
account principles of interaction in social networks.
The purpose: construction of a mathematical apparatus that describes the viral marketing and
identifies its characteristics affecting the development of networks.

2. Materials and research methods

Agent-based modeling, analysis, forecasting, simulation modeling.

3. Results
The multi-agent mathematical model reveals vulnerable vertices and edges of the graph, calculates an
involvement threshold, predicts the development of networks with certain characteristics.
Studies on the phenomenon of information dissemination in social networks is not an easy task, since
connections are often random in nature, the networks are dynamic; it is also difficult to predict the
interest of new users in information dissemination. A prefractal graph serves as a simulation multi-agent
model. The use of a simulation multi-agent approach in modeling epidemics is due to the fact that the

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
HIRM-2019 IOP Publishing
Journal of Physics: Conference Series 1353 (2019) 012122 doi:10.1088/1742-6596/1353/1/012122

approach allows it to take into account various factors influencing the epidemic process and makes it
possible to carry out numerical experiments. A social network is a structure consisting of agents and
relationships (a set of relationships between agents – acquaintance, friendship, cooperation). A social
network is a graph G  (V , E ) where V is the set of vertices (agents) and E is the set of edges describing
the relationship of agents. In the viral marketing in social networks, an agent is a research participant.
Agents are people with individual parameters.
The agent is active, in a certain state which may change under the influence of certain factors.
Properties of the agent include characteristics which form the “level of involvement” u∈ [0,1] of agent
p: numerical - age (v), gender (s), income (d); categorical - marital status (q), education (e), social and
professional criteria (g), reaction speed (b). The “level of engagement” u∈ [0,1] is a parameter indicating
how many people are resistant to advertising. It should be emphasized that one agent may have different
levels of involvement regarding different marketing products; for example, one person may be interested
in cosmetics rather than in car accessories. [1]
Marketing network participants can have two p states:
 active agents p = 1, if the agent accepts an invitation and distributes a marketing virus;
 passive p = 0, if the agent accepts an invitation, but does not distribute a marketing virus.
The position will depend on the level of involvement, since each network participant has an
individual set of characteristics. The model is an artificial city of agents represented as a social network
with the same set of properties different in values, i.e. some agent database. Let us imagine that you
have created a new company and are planning to attract customers through the viral growth. The first
customers, can be friends, etc. The model will have the following inputs:

Variable Description Example

V0 Initial set of clients 8
Number of customers to whom an invitation is sent from one
n 10
customer
The share of invitations that become customers (it depends on
vk 25%
individual parameters of customers who received invitations)

Each network agent may successfully involve a certain number of clients. It is a viral coefficient
K=n·vk [6].
𝑣0 – the vertice of the graph which will correspond to the owner of the company (zero network
member). 𝑣1𝑖 , 𝑖 = ̅̅̅̅̅
1, 𝑛 – vertices representing individual customers who received an invitation from the
customer 𝑣0 , A set of first-level active and passive agents is formed. The process of sending invitations
will be denoted by a directed edge on the graph. Let us connect the vertex 𝑣0 with vertex 𝑣1𝑖 , 𝑖 = ̅̅̅̅̅
1, 𝑛
with directed edges coming from 𝑣0 . In terms of the graph theory, the vertex denoting the primary source
of infection is a root of the marketing virus. Let us combine the resulting set of vertices and edges into
graph 𝐻 = (𝑊, 𝑄):

Figure 1. Graph 𝐻 = (𝑊, 𝑄).

2
HIRM-2019 IOP Publishing
Journal of Physics: Conference Series 1353 (2019) 012122 doi:10.1088/1742-6596/1353/1/012122

Active agents of the first level send invitations to potential members of the network of the second
level. Each vertex of the first level of graph 𝐻 = (𝑊, 𝑄), is connected with 𝑛 vertexes of the second
level 𝑣2𝑖 , 𝑖 = ̅̅̅̅̅̅
1, 𝑛2 . Each first-level agent sends invitations to potential second-level agents [2].
In the resulting graph-model, the number of vertices and edges will increase, but the structure of the
graph which consists of self-similar parts remains stationary. The parts are incentives, and the graph
model is a prefractal graph. Each vertex of the last level is replaced with incentive 𝐻 = (𝑊, 𝑄) –
replacement of the vertex with an incentive (RVI). When repeating the process at l   , the structure
of relations will be a fractal graph G  (V , E ) created by the incentive. There are no ideal cases in practice,
therefore it is necessary to consider a case when the RVI is performed by various incentives
H  H1 , H 2 ,, H s , s Z . [3]
The marketing model will be most appropriate if it takes into account a number of factors:
For social networks, a key indicator is a “viral threshold” (percolation threshold) ñ which is critical
level of involvement above which the marketing virus covers the entire network. If the social network
is represented by a random prefractal graph, the virus with viral coefficient K exponentially multiplies
above the viral threshold; the virus with viral coefficient K exponentially “dies out” below the
percolation threshold. Let us assume that a marketing network agent with an involvement level 𝑢 > ñ
received an invitation from a passive agent; if 𝑢 ≤ ñ the agent will be passive. Let us assign certain
weights to the graph vertices w(vi ), i V , i  1, n , 0  w(vi )  1 ; random numbers are randomly assigned to
the graph vertices by the generator [4].
The percolation theory considers left-to-right or bottom-up processes; we define the concept
“percolation” on the prefractal graph with weighted vertices which are a source and a drain. We assume
that there is a flow between different fixed incentives of the same rank S1  H1  W1 , Q1  and
 
S2  H 2  W2 , Q2  , which are a source and a drain if there is a route u i , v j from vertexes w(v)  c ,
connecting two vertexes where ui  W1 и v j  W2 , u i , v j  I  R , i, j  1, n .

4. Algorithm for searching for a percolation threshold of a prefractal graph

1 1 : 1;
2 The search for different routes connecting the source and the drain at 1 ; if there is such a route,
let us return to step at 1 : 1 2 , otherwise 2 : 1 and let us return to step 3;
3 1 : 1   2 . Если 2  1   ,  : 1 and the algorithm will stop working, otherwise let us
2
return to step 2. [5]
Let us denote the value  as ñ – percolation threshold.
To identify the relationships to be blocked to prevent further the development of the marketing virus,
it is necessary to specify a rank of the incentive and indicate a share of active agent nodes at which
quarantine measures are possible. When the share of active agents reaches the specified rank, all edges
are blocked. These measures may become necessary if the invitation to join the network is harmful; it
may be the dissemination of false information, illegal goods. On the other hand, these edges must be
strengthened, if we need to develop a network.
Let us assume that the model is constructed in the form of a prefractal graph weighted by all vertices.
Let us consider the dissemination of a marketing virus if the primary source of the virus is given. From
a mathematical point of view at the stage 1=1, the structure of distribution of invitations is (n1  1) -vertex
star H1  (W1, Q1 ) .[1]

3
HIRM-2019 IOP Publishing
Journal of Physics: Conference Series 1353 (2019) 012122 doi:10.1088/1742-6596/1353/1/012122

Putting the mathematical description into the verbal correspondence, we will proceed to the next
level r  1 of distribution by the following rules:
1 A lot of incentives H  H1 , H l , H s  , T  2 , build the graph which corresponds to the random
number of infected individuals of the level r ;
2 If the vertex  Vr is not pendant, it is replaced;
3 The replaced vertex  Vr is selected from the subset of pendant vertexes, the edge is incident to
the star center;
4 If any pendant vertex  Vr is not replaced, it is frozen.
Thus, applying the RVI H 2  (W2 , Q2 ) to pendant vertexes, we have graph G2  (V2 , E2 ) , which
was the structure of virus dissemination at the next stage l  2 . Any prefractal graph Gl  (Vl , El ) , l  1,2,...
is a dissemination structure at stage l . Figure 2 shows prefractal graph G3  (V3 , E3 ) , created by a set
of different incentives-stars. Having performed L transitions, we have a root tree GL  (VL , EL ) , which
is a tree of distribution of the infectious disease D L  (VL , EL ) .

Figure 2. Root tree construction.

5. Conclusion
A software simulation product was developed in the FreePascal environment. It is based on the agent
approach that allows for simulation of development of marketing networks, calculation of a viral
threshold, a percolation threshold, choosing measures for strengthening edges that affect the spread of
a marketing virus.

References
[1] Sukharev O S, Kurmanov N V 2017 Model of the social network marketing analysis Bulletin of
the South Russian Technical University. 4 11
[2] Utakaeva I Kh, Kochkarov R A 2010 On the issue of recognition algorithms for the prefractal
graph (NTV SPbGPU. St. Petersburg)
[3] Utakaeva I Kh 2016 Simulation modeling of the spread of epidemics based on the agent approach
Polytematic network electronic scientific journal of the Kuban State Agrarian University 121
1369–1379
[4] Utakaeva I Kh 2011 Mathematical models of infectious dynamics based on prefractal graphs.
Dissertation, Cherkessk
[5] Utakaeva I Kh, Khmelevskaya K A 2017 Econometric modeling using the Python statistical
analysis package Int. Sci. J. 5 67–71
[6] David Skok 2012 Lessons Learned – Viral Marketing. [Electron recourse]. Available at:
https://www.forentrepreneurs.com/lessons-learnt-viral-marketing/