### Abstract

Complex networks are virtually ubiquitous, and the Barabási and Albert model (BA model) has became an acknowledged standard for the modelling of these systems. The so-called BA model is a kind of preferential attachment growth model based on the intuitive premise that popularity is attractive. However, preferential attachment alone is insufficient to describe the diversity of complex networks observed in the real world. In this paper we first use the accuracy of a link prediction method, as a metric for network fitness. The link prediction method predicts the occurrence of links consistent with preferential attachment, the performance of this link prediction scheme is then a natural measure of the ”preferential-attachment-likeness” of a given network. We then propose several modification methods and modified BA models to construct networks which more accurately describe the fitness properties of real networks. We find that all features assortativity, degree distribution and rich-club formation can play significant roles for the network construction and eventual structure. Moreover, link sparsity and the size of a network are key factors for network reconstruction. In addition, we find that the structure of the network which is limited by geographic location (nodes are embedded in a Euclidean space and connectivity is correlated with distances) differs from other typical networks. In social networks, we observe that the high school contact network has similar structure as the friends network and so we speculate that the contact behaviours can reflect real friendships.

Language | English |
---|---|

Pages | 49-60 |

Number of pages | 12 |

Journal | Physica A: Statistical Mechanics and its Applications |

Volume | 474 |

DOIs | |

State | Published - 15 May 2017 |

### Fingerprint

### Cite this

*Physica A: Statistical Mechanics and its Applications*,

*474*, 49-60. DOI: 10.1016/j.physa.2017.01.066

}

*Physica A: Statistical Mechanics and its Applications*, vol 474, pp. 49-60. DOI: 10.1016/j.physa.2017.01.066

**Fitness networks for real world systems via modified preferential attachment.** / Shang, Keke; Small, Michael; Yan, Wei sheng.

Research output: Contribution to journal › Article

TY - JOUR

T1 - Fitness networks for real world systems via modified preferential attachment

AU - Shang,Keke

AU - Small,Michael

AU - Yan,Wei sheng

PY - 2017/5/15

Y1 - 2017/5/15

N2 - Complex networks are virtually ubiquitous, and the Barabási and Albert model (BA model) has became an acknowledged standard for the modelling of these systems. The so-called BA model is a kind of preferential attachment growth model based on the intuitive premise that popularity is attractive. However, preferential attachment alone is insufficient to describe the diversity of complex networks observed in the real world. In this paper we first use the accuracy of a link prediction method, as a metric for network fitness. The link prediction method predicts the occurrence of links consistent with preferential attachment, the performance of this link prediction scheme is then a natural measure of the ”preferential-attachment-likeness” of a given network. We then propose several modification methods and modified BA models to construct networks which more accurately describe the fitness properties of real networks. We find that all features assortativity, degree distribution and rich-club formation can play significant roles for the network construction and eventual structure. Moreover, link sparsity and the size of a network are key factors for network reconstruction. In addition, we find that the structure of the network which is limited by geographic location (nodes are embedded in a Euclidean space and connectivity is correlated with distances) differs from other typical networks. In social networks, we observe that the high school contact network has similar structure as the friends network and so we speculate that the contact behaviours can reflect real friendships.

AB - Complex networks are virtually ubiquitous, and the Barabási and Albert model (BA model) has became an acknowledged standard for the modelling of these systems. The so-called BA model is a kind of preferential attachment growth model based on the intuitive premise that popularity is attractive. However, preferential attachment alone is insufficient to describe the diversity of complex networks observed in the real world. In this paper we first use the accuracy of a link prediction method, as a metric for network fitness. The link prediction method predicts the occurrence of links consistent with preferential attachment, the performance of this link prediction scheme is then a natural measure of the ”preferential-attachment-likeness” of a given network. We then propose several modification methods and modified BA models to construct networks which more accurately describe the fitness properties of real networks. We find that all features assortativity, degree distribution and rich-club formation can play significant roles for the network construction and eventual structure. Moreover, link sparsity and the size of a network are key factors for network reconstruction. In addition, we find that the structure of the network which is limited by geographic location (nodes are embedded in a Euclidean space and connectivity is correlated with distances) differs from other typical networks. In social networks, we observe that the high school contact network has similar structure as the friends network and so we speculate that the contact behaviours can reflect real friendships.

KW - Assortative

KW - BA model

KW - Complex network

KW - Degree distribution

KW - Network sparsity

KW - Rich-club

UR - http://www.scopus.com/inward/record.url?scp=85010189196&partnerID=8YFLogxK

U2 - 10.1016/j.physa.2017.01.066

DO - 10.1016/j.physa.2017.01.066

M3 - Article

VL - 474

SP - 49

EP - 60

JO - PHYSICA A-STATISTICAL MECHANICS AND ITS APPLICATIONS

T2 - PHYSICA A-STATISTICAL MECHANICS AND ITS APPLICATIONS

JF - PHYSICA A-STATISTICAL MECHANICS AND ITS APPLICATIONS

SN - 0378-4371

ER -