Table 1 AUC-PR for triangle prediction on high-school contact data and synthetic hypergraphs from the linear model.

From: Generative hypergraph models and spectral embedding

Data (train:test)

Random

Linear

Arith mean

Geo mean

Harm mean

Highschool (80:20)

5.3e−5

9.3e−4

9.4e−4

6.2e−3

6.0e−3

Highschool (60:40)

1.2e−4

1.2e−3

2.1e−3

1.1e−2

1.1e−2

Highschool (20:80)

2.6e−4

9.8e−3

3.9e−3

7.6e−3

7.4e−3

Primary school (80:20)

3.2e−4

8.4e−3

3.3e−3

1.6e−2

1.7e−2

Primary school (60:40)

1.1e−3

1.2e−2

9.9e−3

2.2e−2

2.2e−2

Primary school (20:80)

1.6e−3

3.0e−2

1.6e−2

2.2e−2

2.1e−2

Linear model (80:20)

6.4e−3

1.6e−1

4.1e−2

7.8e−2

7.7e−2

Linear model (60:40)

1.3e−2

2.8e−1

8.5e−2

1.8e−1

1.8e−1

Linear model (20:80)

3.2e−2

3.0e−1

1.1e−1

1.4e−1

1.3e−1

  1. Highest values are indicated in bold.