### Abstract

We consider the problem of obtaining unbiased estimates of group properties in social networks when using a classifier for node labels. Inference for this problem is complicated by two factors: The network is not known and must be crawled, and even high-performance classifiers provide biased estimates of group proportions. We propose and evaluate AdjustedWalk for addressing this problem. This is a three step procedure which entails: (1) walking the graph starting from an arbitrary node; (2) learning a classifier on the nodes in the walk; and (3) applying a post-hoc adjustment to classification labels. The walk step provides the information necessary to make inferences over the nodes and edges, while the adjustment step corrects for classifier bias in estimating group proportions. This process provides de-biased estimates at the cost of additional variance. We evaluate AdjustedWalk on four tasks: The proportion of nodes belonging to a minority group, the proportion of the minority group among high degree nodes, the proportion of within-group edges, and Coleman’s homophily index. Simulated and empirical graphs show that this procedure performs well compared to optimal baselines in a variety of circumstances, while indicating that variance increases can be large for low-recall classifiers.

Original language | English |
---|---|

Title of host publication | Social Informatics - 10th International Conference, SocInfo 2018, Proceedings |

Editors | Olessia Koltsova, Dmitry I. Ignatov, Steffen Staab |

Publisher | Springer Verlag |

Pages | 67-85 |

Number of pages | 19 |

ISBN (Print) | 9783030011284 |

DOIs | |

Publication status | Published - 1 Jan 2018 |

Event | 10th Conference on Social Informatics, SocInfo 2018 - Saint-Petersburg, Russian Federation Duration: 25 Sep 2018 → 28 Sep 2018 |

### Publication series

Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
---|---|

Volume | 11185 LNCS |

ISSN (Print) | 0302-9743 |

ISSN (Electronic) | 1611-3349 |

### Other

Other | 10th Conference on Social Informatics, SocInfo 2018 |
---|---|

Country | Russian Federation |

City | Saint-Petersburg |

Period | 25/9/18 → 28/9/18 |

### Fingerprint

### Keywords

- Classification error
- Digital demography
- Network sampling
- Quantification learning

### ASJC Scopus subject areas

- Theoretical Computer Science
- Computer Science(all)

### Cite this

*Social Informatics - 10th International Conference, SocInfo 2018, Proceedings*(pp. 67-85). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11185 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-01129-1_5

**Estimating group properties in online social networks with a classifier.** / Berry, George; Sirianni, Antonio; High, Nathan; Kellum, Agrippa; Weber, Ingmar; Macy, Michael.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

*Social Informatics - 10th International Conference, SocInfo 2018, Proceedings.*Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11185 LNCS, Springer Verlag, pp. 67-85, 10th Conference on Social Informatics, SocInfo 2018, Saint-Petersburg, Russian Federation, 25/9/18. https://doi.org/10.1007/978-3-030-01129-1_5

}

TY - GEN

T1 - Estimating group properties in online social networks with a classifier

AU - Berry, George

AU - Sirianni, Antonio

AU - High, Nathan

AU - Kellum, Agrippa

AU - Weber, Ingmar

AU - Macy, Michael

PY - 2018/1/1

Y1 - 2018/1/1

N2 - We consider the problem of obtaining unbiased estimates of group properties in social networks when using a classifier for node labels. Inference for this problem is complicated by two factors: The network is not known and must be crawled, and even high-performance classifiers provide biased estimates of group proportions. We propose and evaluate AdjustedWalk for addressing this problem. This is a three step procedure which entails: (1) walking the graph starting from an arbitrary node; (2) learning a classifier on the nodes in the walk; and (3) applying a post-hoc adjustment to classification labels. The walk step provides the information necessary to make inferences over the nodes and edges, while the adjustment step corrects for classifier bias in estimating group proportions. This process provides de-biased estimates at the cost of additional variance. We evaluate AdjustedWalk on four tasks: The proportion of nodes belonging to a minority group, the proportion of the minority group among high degree nodes, the proportion of within-group edges, and Coleman’s homophily index. Simulated and empirical graphs show that this procedure performs well compared to optimal baselines in a variety of circumstances, while indicating that variance increases can be large for low-recall classifiers.

AB - We consider the problem of obtaining unbiased estimates of group properties in social networks when using a classifier for node labels. Inference for this problem is complicated by two factors: The network is not known and must be crawled, and even high-performance classifiers provide biased estimates of group proportions. We propose and evaluate AdjustedWalk for addressing this problem. This is a three step procedure which entails: (1) walking the graph starting from an arbitrary node; (2) learning a classifier on the nodes in the walk; and (3) applying a post-hoc adjustment to classification labels. The walk step provides the information necessary to make inferences over the nodes and edges, while the adjustment step corrects for classifier bias in estimating group proportions. This process provides de-biased estimates at the cost of additional variance. We evaluate AdjustedWalk on four tasks: The proportion of nodes belonging to a minority group, the proportion of the minority group among high degree nodes, the proportion of within-group edges, and Coleman’s homophily index. Simulated and empirical graphs show that this procedure performs well compared to optimal baselines in a variety of circumstances, while indicating that variance increases can be large for low-recall classifiers.

KW - Classification error

KW - Digital demography

KW - Network sampling

KW - Quantification learning

UR - http://www.scopus.com/inward/record.url?scp=85057321263&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85057321263&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-01129-1_5

DO - 10.1007/978-3-030-01129-1_5

M3 - Conference contribution

SN - 9783030011284

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 67

EP - 85

BT - Social Informatics - 10th International Conference, SocInfo 2018, Proceedings

A2 - Koltsova, Olessia

A2 - Ignatov, Dmitry I.

A2 - Staab, Steffen

PB - Springer Verlag

ER -