A Survey on Tie Strength Estimation Methods

in Online Social Networks

Isidoros Perikos

and Loizos Michael

1,2

Open University of Cyprus, Nicosia, Cyprus

CYENS Center of Excellence, Nicosia, Cyprus

Keywords: Tie Strength, Social Networks, Social Relations, Social Interaction.

Abstract: Social networks constitute an important medium for social interaction where people communicate and

formulate relationships in a way similar to what they do in real life. The analysis of the users’ relationships

in social networks can lead to new insights into human social behavior. Tie strength constitutes a core

aspect of social relationships, which represents the importance of a relationship and the closeness of

individuals. Understanding the key features of tie strength in social networks can assist in formulating more

efficient user-centric services. This survey paper examines the advances in the area of the analysis of tie

strength in social networks. We study the dimensions of tie strength and review the key predictive features

for each dimension. We, then, undertake a comparative study of methodologies to model tie strength and

examine the key findings. Finally, we discuss open issues and challenges in specifying tie strength.

1 INTRODUCTION

During the last years, the advancement of social

networks has completely redefined the way that we

conceive our social relationships and has created the

sensation of having broken the barriers of time and

geography that are limiting people’s social world

(Pappalardo et al., 2012). With the rising expansion

of the social networks the capacity of the people to

interact, communicate and network has been greatly

increased (Liberatore and Quijano-Sanchez, 2017).

Social networks create new online environments

where social relationships can not only map and

develop preexisting relationships that are established

face-to-face in the physical world, but can also

facilitate the development of new relationships that

may exist and evolve only in the world of the social

network (Arnaboldi et al., 2015).

Social networks constitute environments where

social ties among individuals are developed and

have become a predominant medium for social

interaction that have changed completely the way of

human communication. People connect and

formulate relationships in social media in a similar

way they do in the real world (Dunbar et al., 2015).

Actually, the formulation and the development of

social relationships is actually what makes social

media ‘social’ (Gilbert and Karahalios, 2009).

Although individuals have different types of

relationships with other ones that differ in kind and

closeness, social networks do not distinguish and

diversify them. So, they treat all connections the

same, even if one relationship refers to a trusted

friend and another to a total stranger. In social

networks, the online connection between users is

often defined as a ‘friendship’ or ‘follow’ or

‘connected’ manifestation and a connection between

two individuals is considered to either exist or not. So,

different types of friendships best friends, total

strangers and long acquaintances are all grouped

under the label of ‘friend’, ‘follow’ or ‘connection’. In

this regard, social networks do not make distinctions

between best friends, that are the relationships that

one trusts, and mere acquaintances and so all the

relationships are uniformly labeled. However, some

social relationships and connections are stronger than

others. It is natural for people to have not only friends

but also best friends and also to distinguish friends

from acquaintances (Jones et al., 2013). Social

scientists highlighted this point of the social

connections and researchers utilize the expression of

the tie strength to pertain to this concept (Granovetter,

1973; Marsden and Campbell 2012).

Tie strength is a key concept in social networks

associated with the value which is placed by

individuals on their relationships referring to the

484

Perikos, I. and Michael, L.

A Survey on Tie Strength Estimation Methods in Online Social Networks.

DOI: 10.5220/0010845100003116

In Proceedings of the 14th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2022) - Volume 3, pages 484-491

ISBN: 978-989-758-547-0; ISSN: 2184-433X

general sense of closeness with another individual

(Granovetter, 1983). So, relationships in social

networks can be measured with the currency of the tie

strength (Gilbert and Karahalios, 2014), a concept

introduced by Granovetter (1973) which has become

the metric to measure the relationships in social

media. In this context, two types of ties are specified,

strong ties and weak. In general, when the sense of

closeness between two individuals is strong, a strong

tie is defined and, in the same regard, when it is weak,

a weak tie is defined (Granovetter, 1983). Strong ties

are considered to exist with people one trusts, and that

their social circles can highly overlap with his own.

Weak ties are mainly considered acquaintances.

Measuring and predicting tie strength, and

moreover, understanding the factors that drive tie

strength, has been an expanding area of interest in

social sciences, with increasing utility in the analysis

of social networks (Mattie al., 2018). Analyzing and

predicting tie strength in social networks can lead to

new insights into human social behavior and assist

in designing novel user-centric services (Arnaboldi

et al., 2013). So, the analysis of the social

relationships and the accurate specification of the tie

strength is highly desired.

In the context of this work, we examine the

advent of the last decade in predicting tie strength in

social networks. The survey reviews the elements

and the characteristics of tie strength and categorizes

the research works with respect to the dimensions

that are involved in the modelling of the

relationships’ tie strength. The reminder of the paper

is structured as follows. Section 2 examines the

dimensions and manifestations of tie strength in

social networks while Section 3 examines the

predictive variables that can be extracted from

interaction data and associates the variables to

dimensions. Section 4 examines prediction methods

and models for tie strength estimation, categorizes

them according to the dimension of tie strength that

are considered and reviews the main findings

regarding the components of the tie strength and the

good predictors. Section 5 examines the utilization

in a wide spectrum of social network analyses.

Section 6 examines challenges and open research

directions. Finally, Section 7 concludes the work.

2 DIMENSIONS OF TIE

STRENGTH

Tie strength constitutes a factor of high importance

in the analysis of social networks and it is

considered to be a complex factor that is hard to

accurately estimate. The main reason for this is that

tie strength is a multidimensional factor where

different forms and levels of interaction need to be

considered. Tie strength is highlighted to have many

dimensions and different manifestations.

Granovetter (1973), in his landmark work on The

Strength of Ties, specified four main dimensions for

tie strength: the time spent connecting and

interacting with others, the emotional intimacy, the

intensity and the reciprocal services. After that, three

additional dimensions were proposed and the list of

dimensions was extended with the proposal of the

emotional support (Wellman and Wortley, 1990) the

social distance (Lin et al., 1981), and the structural

topology of the social network (Xiang et al., 2010).

Each dimension captures different elements of the

social relationships. The dimension of time captures

the duration and the frequency of the

communication. In general, the more frequent and

higher the interaction between a pair of individuals

is, the stronger the sentiment of friendship and tie

people feel (Luarn and Chiu, 2015; Mathews et al.,

1998). Strong tie is bound up with the constant and

frequent communication and the amount of time can

promote other dimensions too (He et al., 2012).

The intensity represents the recognition of entities

producing emotions that stresses on the cognition of

others. It is relative to the absolute strength and

individuals with highly intensive relationships is

expected to spend more time with each other, that is

greatly more than individuals with relationships that

are less intensive (Kwon et al., 2013).

The intimacy concerns the affection between two

individuals and acts as a sense of security and

reliance (He et al., 2012). It is stated that intimacy

relationships are willing to talk with open mind and

demonstrate great support and recognition. It

necessitates considerably more commitment and

presumably a greater amount of positive affect

between each other (Lewis et al., 2008)

The dimension of the reciprocal services

represents the different forms of communication and

the services utilized in interaction. An important

parameter to develop a relationship is revenue that

can be measured by the cost and the profit including

energy, time emotion, and others. Social networks

can reduce the cost of the social activities (He et al.,

2012). Strong ties can easily share the information

and the resources they possess and also they can

provide access to information circulating in their

dense network. So, strong tie includes more

reciprocity services in exchanges (Granovetter,

1983).

A Survey on Tie Strength Estimation Methods in Online Social Networks

485

The emotional support dimension represents a tie

on an emotional level and concerns cases of

discussions and advice offering on personal and

family problems, something that can indicate a

strong tie between the users (Gilbert and Karahalios,

2009). The dimension refers to providing messages

that involve emotional content, re-assuring that the

one is valuable and care about. Strong ties provide

powerful emotional support that unites to face

challenges and overcome crises.

Figure 1: Overlap of the four main dimensions defined by

Granovetter (1973).

Structural dimension represents factors such as

the topology of the social network and social circles

of the users in it (Xiang et al., 2010). Strong ties and

more likely to connect similar people and similar

individuals tend and are more likely to cluster

together. So, given that strong ties connect

individual A to B and also to C, then it is very likely

that C and B will develop a friendship once they

meet (Granovetter, 1983).

Social distance is also highlighted to influence tie

strength and factors like gender, race,

socioeconomic status, education and political views

and affiliation and can affect the tie strength

development between individuals. Research studies

have indicated that strong ties are more common

between individuals of the same age, interests and

who share certain life activities (Gilbert and

Karahalios, 2009; Luarn and Chiu, 2015).

3 PREDICTIVE VARIABLES FOR

TIE STRENGTH

The dimensions have facilitated the definition and

the quantification of possible factors and predictive

variables of tie strength (Mattie et al., 2018). These

variables derive from the social information in the

networks that relate to the profiles of individuals as

well as to the interaction with their peers and which

will be used as predictors of tie strength between

two individuals. Table 1 categorizes the predictive

variables used in research works in the literature.

The predictive variables are mapped into the seven

dimensions of the tie strength. Given that different

social networks provide their users with different

means of interaction, some variables generalize to

any network, while some other may be specific to a

number of social networks (Mattie et al., 2018). For

example, photo tags variables and check-in denoting

that individuals appear together in photos. User

social profiles and interaction activities with their

peers in a social network need to be analyzed in

order to identify relative variables that can be

quantified and be used in order to infer tie strength

between individuals.

In the dimensions of time, time since first

communication measures the length of the connection

while the time since last communication captures the

recency. The frequency is a proxy for the volume of

the interaction between two individuals.

In the dimension of intensity, exact

communication aspects and messages are measured

like the number of the messages exchanged, the

posts, comments, likes. The variables of this

dimension rely heavily on the characteristics and the

communication means of each social network.

In the intimacy dimension, intimacy words

measure the topics of the messages exchanged while

the relationship status captures specific types of

relationships that may be denoted by the users such as

married with each other, family members, etc.

Common appearances in photos is another

ICAART 2022 - 14th International Conference on Agents and Artiﬁcial Intelligence

486

measurement of the intimacy counting photographs

that the two individuals appear together and the

common check-ins measures places that they have

been together.

In the dimension of reciprocal services, common

applications measure the services and the applications

that both the user and the friend share. The same

stands for the links exchanged where URLs passed

between two users can be indicative of the reciprocal

services both use (Gilbert and Karahalios, 2009).

In the dimension of structural topology, variables

capture aspects of the network structure and the

groups that the individuals belong to. So, common

groups variable measures the groups that the two

individuals belong to, while the overlapping networks

Table 1: Categorization of variables.

Dimension Main Predictive Variables

Time

Time since first communication

Time since last communication

Frequency of communication

Intensity

Communication aspects with frien

Messages exchanged

Comments exchan

Likes

Intimacy

Relationship status

Appearances together in photos

Common check-ins

Intimacy words in the

communication

Reciprocal

Services

Common applications

Links exchanged by wall posts

Structural

Topology

Common groups

Mutual friends

TF-IDF of interests

Listed in overlapping networks

Betweeness

Centrality

Emotional

Support

Text analysis and specification of

emotional context and emotional

words. Specification of positive

and negative contexts

Social

Distance

Age difference

Occupation difference

Education difference

Political difference

Religion difference

capture the social circles, organizations and networks

like universities and companies that both individuals

are members. Mutual friends can also indicate clues

for tie strength and having mutual friends can foster

relationship development (Adamic and Adar, 2003).

Structural variables can be measured by the interests

individuals have in common and the normalized TF-

IDF of the interests too.

In the dimension of emotional support, deep

analysis of text messages and interaction of the

individuals aims to specify emotional support

indicators and predictive variables like positive

emotional words and negative emotional words.

Dictionaries and linguistic resources like LIWC can

provide indicative information about the categories

of the words and the context messages.

In the social distance dimension, variables

measure the age difference, the education discipline

and level of the individuals, the political view and

the occupation status. The identity information of

the profiles as well as the language (location) beliefs

(philosophy, political view) are used to measure the

social distance of the individuals (Luarn and Chiu,

2015).

4 METHODS FOR MODELLING

TIE STRENGTH

In the literature, several works study tie strength

with the aim to formulate predictive methods to

estimate it. The first direction for the estimation of

tie strength was indicated by Granovetter (1973)

who stated that tie strength can indeed be quantified

and that the strength of a tie is a probably linear

combination of the amount of time, the intimacy, the

intensity and the reciprocal services (Granovetter,

1973). During the last years, research works have

examined different methods based on the

characteristics of the particular research domain

examined, the aims of the study and the access to the

predictive variables that can be calculated in the

domain examined. Typically, users are involved and

participate in surveys in order to label the nature of

each one of their social relationships.

Methods to obtain the ground truth data of the tie

strength of relationships in a social network is a

process of special attention. In the literature, two

main stream of approaches prevail. The first

approach which is the most common is to survey the

users who accepted to participate in the study and

collect feedback about the strength of their

relationships with their friends (Jones et al., 2013).

A Survey on Tie Strength Estimation Methods in Online Social Networks

487

The second approach is to use trusted networks in

order to determine strong ties (Rotabi et al., 2017).

An example is the use of telephone network which is

considered to be a trusted network and the tie

strength between two individuals who have phone

contacts is determined as strong one.

Kahanda and Neville (2008) examined the nature

and the dimensions of the relationship strength in

Facebook and utilized a set of characteristics like

marital status, gender, topological features like user

connectivity, graph of friendship, shared posts on the

wall to specify special friends. The authors applied a

supervised learning approach and the results of the

study concluded that the network transactional

features like shared posts on the wall are the most

prominent in predicting tie strength.

Gilbert and Karahalios (2009) defined indicators

of tie strength that are specific to Facebook users

and formulate a regression predictive model which

reports 85% accuracy for the classicization of binary

tie strength. The authors conclude that the dimension

of intimacy makes the greatest contribution to tie

strength specification and that educational difference

strongly assists in the prediction of tie strength with

tie strength diminishing as difference grows.

Pappalardo et al. (2012) present a model to

estimate tie strength that is based strongly on

variables of the network topology dimension and the

intensity. The authors describe a quantitative

measure of tie strength that is domain-independent

and can be generalized and applied to any social

network. The findings indicate that the strength of a

tie is strictly related to the number of interactions

among the people involved and that it is also related

to the number of different contexts in which those

connections take place.

Arnaboldi et al. (2013) present a linear model

which estimates tie strength in Facebook. The

authors’ model takes into account variables from the

dimensions of time, network structure, intensity and

intimacy and describing different aspects of user

interaction. The liner model reports quite good

performance and has accuracy higher than 80%. The

work indicates that the recency of contact is the most

indicative predictor of tie strength.

Jones et al. (2013) determine tie strength from

users’ behavior in Facebook. The authors’ model

takes into account variables from the dimensions of

time,

intimacy, reciprocal services, intensity, structure

topology and social distance. The authors in their

study surveyed Facebook users asking them to

specify their closest friends, a piece of information

that is used as ground truth. An additive logistic

regression model was formulated which achieved an

accuracy of 84% on the context of the study. The

authors report the frequency of online interaction

was the most indicative information for strong ties.

Servia-Rodriguez et al. (2014) present a model to

assess strength and classify it within four categories

of social spheres. The model assesses tie strength by

taking into account users’ interactions and predictive

features from the dimensions of intensity, reciprocal

services, intimacy and structural topology. The work

also points out the importance of using information

from as many social networks as possible in order to

avoid losing data in estimation of the tie strength.

Table 2: Analysis of works with respect to the dimensions.

Work Time Intensity Intimacy Reciprocal

Services

Emotional

Suppor

Structure

topolo

Social

Distance

(Granovetter, 1973) x x x x

(Kahanda and Neville,

2008)

x x x

(Gilbert and Karahalios

2009)

x x x x x x

(Pap

alardo et al., 2012) x x

(Arnaboldi et al., 2013) x x x x

(Jones et al., 2013) x x x x x x

(Servia-Rodriguez et al.,

2014)

x x x

(Liberatore and Quijano-

Sanchez, 2017)

x x x x x x x

(Mattie et al., 2018) x x x

(Stolz and Schlereth,

2020)

x x

(Ureña-Carrion, 2020) x x x x x

ICAART 2022 - 14th International Conference on Agents and Artiﬁcial Intelligence

488

The reason for this is the fact that people usually

possess accounts in different social platforms and it

is possible not to interact with their peers with the

same frequency in all of them.

Liberatore and Quijano-Sanchez (2017) present a

computational model for tie strength that is

independent of the domain of the social network. In

the context of the work, users were requested to

participate and define their relationships. The

authors analyzed their personal and friendship

interaction data and a linear model was examined.

Predictive variables from the seven dimensions were

examined and authors in their study indicate that

relying exclusively and solely on intensity or

intimacy may not be enough to efficiently calculate

the tie strength.

Mattie et al. (2018) present a method for tie

strength estimation that is based mainly on the

structure dimension. The bow tie framework is

proposed which consists of a focal tie and all actors

connected to either or both of the two focal nodes on

either side of the focal tie. The authors utilize

machine learning techniques as well as regression

methods and study which variables are most useful

in predicting tie strength. The results show that the

more the friend that two individuals share the

stronger their tie is and also that geographical

location can increase the tie strength of individuals.

Stolz and Schlereth (2020) present an approach

for the prediction of tie strength that takes into

account predictive variables from network, social

distance and intensity dimensions. The proposed

approach relies heavily on ego network structures

that are user connections and interlinkage among

them. The authors include also user similarity

variables such as the matching gender and language

of the users. The precision of the authors’ method in

identifying all observed strong ties is 45%. The work

also indicates that individuals react stronger to

suggestions that are made of a close friend compared

to the suggestions made by an acquaintance.

Ureña-Carrion et al. (2020) study how

communication events and contact time can be an

indicative predictor for tie strength. The authors

study tie strength through the four main dimensions

of Granovetter’s theory. The results of the authors’

work indicate that the number of days and hours

with contracts are quite indicative variables in the

estimation of tie strength. Also, the time of the first

and last communication can provide indicative

information of tie strength and perform better than

the communication-intensity variables.

5 IMPACT OF TIE STRENGTH

ON SOCIAL ANALYSES

METHODS

Tie Strength is a prevalent feature in social network

research and various studies are examining the

impact of tie strength in various procedures.

Understanding of the tie strength is essential in order

to study the dynamics of social behaviors in network

as well as the relationships of the users in it

(Arnaboldi et al., 2013). Tie strength can to affect

career advancement and the word-to-mouth

propagation of information (Mittal et al., 2008). It is

highlighted that the tie strength has a general impact

on behavior outcomes and intentions in the social

network contexts (Ureña-Carrion et al., 2020).

Specifically, strong ties are the ones that are most

likely to transit norms and behavior change (Kim et

al., 2015) and the identification of strong ties in a

social network can assist in focus targeting these

relationships in a wide spectrum of positive

interventions that can have great multiplier effects as

they spread from an individual to another (Jones et

al., 2013). Research studies have indicated the

impact of tie strength on decision making and that

strong ties can greatly affect users in opinion

seeking as well as in adoptions of stances (Stolz and

Schlereth, 2020). The information about the tie

strength and the knowledge of the social dynamics

that affect and contribute to tie strength is reported

to increase the efficiency of link prediction in social

networks (Mattie et al., 2018). Tie strength

estimation in the context of online social networks

can assist in more efficiently detect communities in

social networks. Tie strength is also assistive in

modeling information diffusion in social networks

(Bakshy et al., 2012). The literature suggests that

novel information comes mainly from weak ties.

Weak ties can provide access to novel information

that is pieces of information that are not circulating

in the dense network formulated by strong ties

(Gilbert and Karahalios, 2009). Weak ties might be

more frequently to convey new pieces of

information, so even though someone is somebody

you don't interact much with, and you might have

less in common with, they might connect you to a

part of the world that you don't normally have access

to, and so they can still be very important in

accessing information that might not be redundant

with people that you often interact with. So, weak

ties can play a very important role in information

diffusion and to provide access to novel pieces of

A Survey on Tie Strength Estimation Methods in Online Social Networks

489

information. Moreover, of importance is also the

bandwidth that a connection possesses.

6 CHALLENGES AND OPEN

ISSUES

Although the methods in the literature achieve quite

promising performance in predicting tie strength

there are many challenges and open issues to be

addressed. A first challenge concerns the detection

and the proper handling of salient users. Most of the

methods in the literature merely focus on active

users and the salient users will be inappropriately

considered as acquaintances due to their inactiveness

(Li et al., 2018).

Another challenge concerns the emotional

support dimension of tie strength and the fact that

emotional factors are not directly measurable and are

quite hard to quantify (Arnaboldi et al., 2013;

Gilbert and Karahalios, 2009). The more efficient

measurement of indicative variables from the

emotional dimension could greater impact tie

strength prediction.

The social distance dimension needs special

handling too. An analytic framework for the fine-

grained measurement of the social distance. Most of

the existing methods do not properly capture the

social distance as it derives from differences in

education, in political views and so the specification

of an analytical and fine-grained framework could

be assistive in measuring social distances and reflect

the diversity in the population.

Furthermore, a fine-grained estimation of the tie

strength will also be a next. Most of the existing

works model and estimate tie strength on a binary

scale where strong and weak ties are specified

between the users in a social network. Although

research studies of the previous years have provided

advances and valuable insights into the dimensions

and the predictive variables, the analytical and

numerical specification of tie strength constitutes a

direction for research (Mattie et al., 2018).

Reproducibility issues constitute a great challenge

too since most of the approaches in the literature

utilize so many features and predictive variables from

social networks (Facebook, Twitter) that their models

are almost impossible to reproduce mainly because of

recent APIs restrictions (Liberatore and Quijano-

Sanchez, 2017). In addition, in the literature, there is a

lack of public benchmark data sets for the study of

methods and for systematically appraising the

performances of models. The availability of such

benchmark, pseudonymized or anonymized datasets

for the fair assessment of new methods and models is

a next important step too.

Kin relationships need special attention too.

Research studies indicate that kin relationship

remain stable over time even if they interact rarely

or even not at all (Roberts and Dunbar, 2011). So,

such relationships need special handling in modeling

user interactions and formulating models for tie

strength prediction. The proper identification of kin

relationships and the formulation of social models to

handle them, could further enhance the performance

of prediction models and provides another direction

for further research.

7 CONCLUSIONS

Tie strength constitutes a primitive challenge in the

domain of social networks and the modeling and

prediction of the tie strength of users’ relationships

has attracted increased research interest. This paper

presents the advent of the last years in predicting tie

strength in social networks and examines methods

for the estimation of tie strength. The dimensions

and the manifestations of the tie strength are studied

and state-of-the-art research works in tie strength

estimation are examined with respect to the different

tie strength dimensions. Meta-analyses on the results

and the findings of the studies in the literature are

performed. Last but not least, main challenges and

open issues in designing methods for the estimation

of tie strength are discussed.

ACKNOWLEDGEMENTS

This work was supported by funding from the EU's

Horizon 2020 Research and Innovation Programme

under grant agreements no. 739578 and no. 823783,

and from the Government of the Republic of Cyprus

through the Deputy Ministry of Research,

Innovation, and Digital Policy.

REFERENCES

Adamic, L. & Adar, E. (2003) Friends and Neighbors on the

Web. Social Networks, 25(3):211–230.

Arnaboldi, V., Passarella, A., Conti, M. & Dunbar, R.

(2015) Online Social Networks: Human Cognitive

Constraints in Facebook and Twitter Personal Graphs.

Elsevier.

ICAART 2022 - 14th International Conference on Agents and Artiﬁcial Intelligence

490

Arnaboldi, V., Guazzini, A. & Passarella, A. (2013)

Egocentric Online Social Networks: Analysis of Key

Features and Prediction of Tie Strength in Facebook.

Computer Communications, 36(10–11):1130–1144.

Bakshy, E., Rosenn, I., Marlow, C. & Adamic, L. (2013)

The Role of Social Networks in Information Diffusion.

In Proceedings of the 21st International Conference on

World Wide Web, pp. 519–528.

Dunbar, R., Arnaboldi, V., Conti, M. & Passarella, A.

(2015) The Structure of Online Social Networks Mirrors

Those in the Offline World. Social Networks, 43:39–47.

Flamino, J., DeVito, R., Szymanski, B. K. & Lizardo, O.

(2021) A Machine Learning Approach to Predicting

Continuous Tie Strengths. arXiv preprint

arXiv:2101.09417.

Granovetter, M. (1983) The Strength of Weak Ties: A

Network Theory Revisited. Sociological Theory, 1:201–

233.

Granovetter, M. (1973) The Strength of Weak Ties. The

American Journal of Sociology, 78(6), 1360–1380.

Gilbert, E. & Karahalios, K. (2009) Predicting Tie Strength

with Social Media. In Proceedings of the SIGCHI

Conference on Human Factors in Computing Systems,

pp. 211–220.

Gilbert, E. & Karahalios, K. (2014) Computing and

Building around Tie Strength in Social Media.

Foundations and Trends in Human-Computer

Interaction, 7(3):237–349.

He, Y., Zhang, C. & Ji, Y. (2012) Principle Features for Tie

Strength Estimation in Micro-Blog Social Network. In

Proceedings of the 12th IEEE International Conference

on Computer and Information Technology, pp. 359–

367.

Kahanda, I. & Neville, J. (2009) Using Transactional

Information to Predict Link Strength in Online Social

Networks. In Proceedings of the AAAI International

Conference on Web and Social Media.

Khan, N. A., Zhang, S., Zhou, W., Almogren, A., Ud, D. &

Asif, M. (2020), Inferring Ties in Social IoT Using

Location-Based Networks and Identification of Hidden

Suspicious Ties. Scientific Programming.

Kim, D., Hwong, A., Stafford, D., Hughes, A., O'Malley, J.,

Fowler, J. & Christakis, N., (2015) Social Network

Targeting to Maximise Population Behaviour Change:

A Cluster Randomised Controlled Trial. The Lancet,

386(9989):145–153.

Kwon, O., Kim, C. & Kim, G. (2013) Factors Affecting the

Intensity of Emotional Expressions in Mobile

Communications. Online Information Review.

Jones, J., Settle, J., Bond, R., Fariss, C., Marlow, C. &

Fowler, J. (2013) Inferring Tie Strength from Online

Directed Behavior. PloS One, 8(1):e52168.

Larson, J. (2017) The Weakness of Weak Ties for Novel

Information Diffusion. Applied Network Science,

2(1):1–15.

Lewis, K., Kaufman, J., Gonzalez, M., Wimmer, A. &

Christakis, N. (2008) Tastes, Ties, and Time: A New

Social Network Dataset using Facebook.com. Social

Networks, 30(4):330–342.

Li, X., Fang, H., Yang, Q. & Zhang, J. (2018) Who Is Your

Best Friend? Ranking Social Network Friends

According to Trust Relationship. In Proceedings of the

26th Conference on User Modeling, Adaptation and

Personalization, pp. 301–309.

Liberatore, F. & Quijano-Sanchez, L. (2017) What Do We

Really Need to Compute the Tie Strength? An

Empirical Study Applied to Social Networks. Computer

Communications, 110:59–74.

Lin, N., Ensel, W. & Vaughn, J. (1981) Social Resources

and Strength of Ties: Structural Factors in Occupational

Status Attainment. American Sociological Review,

46(4):393–405.

Luarn, P. & Chiu, Y. P. (2015) Key Variables to Predict Tie

Strength on Social Network Sites. Internet Research,

25(2):218–238.

Marsden, P. & Campbell, K. (2012) Reflections on

Conceptualizing and Measuring Tie Strength. Social

Forces, 91(1):17–23.

Mathews, K. M., White, M. C., Long, R. G., Soper, B. &

Von Bergen, C. W. (1998) Association of Indicators and

Predictors of Tie Strength. Psychological Reports,

83:1459–1469.

Mattie, H., Engø-Monsen, K., Ling, R. & Onnela, P. (2018)

Understanding Tie Strength in Social Networks Using a

Local Bow Tie Framework. Scientific Reports, 8(1):1–

Mittal, V., Huppertz, J. & Khare, A. (2008) Customer

Complaining: The Role of Tie Strength and Information

Control. Journal of Retailing, 84(2):195–204.

Pappalardo, L., Rossetti, G. & Pedreschi, D. (2012) How

Well Do we Know Each Other? Detecting Tie Strength

in Multidimensional Social Networks. In Proceedings of

the IEEE/ACM International Conference on Advances

in Social Networks Analysis and Mining, pp. 1040–

1045.

Rotabi, R., Kamath, K., Kleinberg, J. & Sharma, A. (2017)

Detecting Strong Ties Using Network Motifs. In

Proceedings of the 26th International Conference on

World Wide Web, pp. 983–992.

Roberts, S. & Dunbar, R. (2011) The Costs of Family and

Friends: An 18-Month Longitudinal Study of

Relationship Maintenance and Decay. Evolution and

Human Behavior, 32(3):186–197.

Servia, R., Diaz, R., Fernandez, V., Blanco, F. & Pazos, A.

(2014) A Tie Strength Based Model to Socially-

Enhance Applications and its Enabling Implementation:

mySocialSphere. Expert Systems with Applications,

41(5):2582–2594.

Stolz, S. & Schlereth, C. (2020) Predicting Tie Strength

with Ego Network Structures. Journal of Interactive

Marketing, 54:40–52.

Ureña-Carrion, J., Saramäki, J. & Kivelä, M. (2020)

Estimating Tie Strength in Social Networks Using

Temporal Communication Data. EPJ Data Science,

9(1):no.37.

Wellman, B. & Wortley, S. (1990) Different Strokes From

Different Folks: Community Ties and Social Support.

American Journal of Sociology, 96(3):558–588.

Xiang, R., Neville, J. & Rogati, M. (2010) Modeling

Relationship Strength in Online Social Networks. In

Proceedings of the 19th International Conference on

World Wide Web, pp. 981–990.

A Survey on Tie Strength Estimation Methods in Online Social Networks

491