Synthetic Multi-field Networks Generated by SASCA-ReS

The networks in this repository were created by simulation under the SASCA-ReS model (more information can be found in the <a href="https://github.com/illinois-or-research-analytics/SASCA-ReS">Github link</a> here or below). SASCA-ReS is a generative model of citation that takes an input citation network along with tabular data for the frequencies of out_degree (citations made) and recency and returns a larger citation network based on user-specified parameters. The networks are used as benchmarks for clustering software and are roughly 14M, 100M, and 272M nodes in size respectively. What the field names in the there abm* networks mean is explained in detail on the SASCA-ReS Github site and the published article. The data types in the parquet files are listed below.

For the simulations that generated the three networks in this repository (abm14mf, abm100mf, and abm272mf), a multi-field input network was constructed by performing a one-hop breadth-first search (BFS) on five articles using the Dimensions database, one each from the field of mathematics, physics, chemistry, biology, and scientometrics. Each of these articles was published between 1995 and 1998 and is well-cited. The total number of articles harvested was 26,595 with nine articles appearing in two fields (~0.03%). For these nine, a single field was assigned by randomly selecting from one of the two labels resulting in a final count of articles across fields as biology (13,625), physics (6,549), chemistry (4,015), scientometrics (1,263), mathematics (1,235). This network with its 26,595 nodes and 272,636 edges after filtering for temporal consistency (citing year ≥ cited year) was used as input to simulations under SASCA-ReS. The multi-field input network is also included in the repo as simready_nodelist.csv and simready_edgelist.csv. Two additional files derived from the multi-field input network and used in these simulations are also included. (i) the out_degree frequency -> multifield_out_degree_bag.csv (ii) recency frequency -> multifield_reccount.csv.

abm14mf: 13,558,804 nodes; 256,814,390 edges
abm100mf: 100,285,739 nodes; 1,901,039,159 edges
abm272mf: 272,739,483 nodes; 5,170,223,924 edges

Nodelist data types:
type: string
year: int32
alpha: double
pa_weight: double
fit_weight: double
fit_lag_duration: int32
fit_peak_value: int32
fit_peak_duration: int32
in_degree: int32
out_degree: int32
assigned_out_degree: int32
planted_nodes_line_number: int32
generator_node_string: string
sampled_neighborhood_size: int32
fully_random_citations: int32

Edgelist data types:
source: int32
target: int32