Network Community Partition

This is a catalogue and repository of network datasets with the aim of aiding scientific research. This website is meant to be browsed both by humans and machines alike, and can also be accessed via a convenient JSON API, or via the graph-tool library. The network datasets themselves are available in several machine-readable formats, in particular gt, GraphML, GML and CSV. The upstream origin of each dataset is meant to be as transparent as possible. Each dataset contains its own publicly available extraction and parsing script, accessible via a git repository, which also includes the entire code for this website, released as Free Software under the AGPLv3. Users are encouraged to inspect the entire pipeline from original upstream data publication, downloading, parsing and format conversion. Users are also welcome to report problems or omissions with the datasets, as well as suggest new ones, either by opening an issue, or simply by forking the git repository and proposing a merge request.

This dataset is a collection of undirected and unweighted LFR benchmark graphs as proposed by Lancichinetti et al. [1]. We generated the graphs using the code provided by Santo Fortunato on his personal website [2], embedded in our evaluation framework [3], with two different parameter sets. Let N denote the number of vertices in the network, then Maximum community size: 0.2N (Set A); 0.1N (Set B) Minimum community size: 0.05N (Set A); 10 (Set B) Maximum node degree: 0.19N (Set A); 0.19N (Set B) Community size distribution exponent: 1.0 (Set A); 1.0 (Set B) Degree distribution exponent: 2.0 (Set A); 2.0 (Set B). All other parameters assume default values. We provide graphs with different combinations of average degree, network size and mixing parameter for the given parameter sets: Set A: For average degrees in {15, 25, 50} we provide network sizes in {300, 600, 1200}, each with 20 different mixing parameters linearly spaced in [0.2, 0.8]. For each configuration we provide 100 benchmark graphs. Set A: For average degrees in {15, 25, 50} we provide mixing parameters in {0.35, 0.45, 0.55}, each with network sizes in {300, 450, 600, 900, 1200, 1800, 2400, 3600, 4800, 6200, 9600, 19200}. For each configuration we provide 50 benchmark graphs. Set B: For average degrees in {20} we provide network sizes in {300, 600, 1200, 2400}, each with 20 different mixing parameters linearly spaced in [0.2, 0.8]. For each configuration we provide 100 benchmark graphs. Benchmark graphs are given in edge list format. Further, for each benchmark graph we provide ground truth communities as membership list and as structured datatype (.json), its generating random seeds and basic network statistics. [1] Lancichinetti A, Fortunato S, Radicchi F (2008) Benchmark graphs for testing community detection algorithms. Physical Review E 78(4):046110,https://doi.org/10.1103/PhysRevE.78.046110 [2] https://www.santofortunato.net/resources, Accessed: 19 Jan 2021 [3] https://github.com/synwalk/synwalk-analysis, Accessed: 19 Jan 2021

The **Perfume Co-Preference Network ** dataset comprises comprehensive user reviews and ratings collected from the Persian retail platform Atrafshan. This dataset, central to our research on community detection in fragrance preferences, includes 36,434 comments from 7,387 unique users, providing insights into consumer sentiment towards various perfumes. It is designed to facilitate the analysis of user preferences through sentiment analysis, allowing for the clustering of perfumes based on shared attributes. The dataset features three main components: 1. User Reviews and Perfume Attributes Dataset: This captures user sentiments expressed in comments, along with metadata such as user IDs, perfume details, and ratings across key attributes (scent, longevity, sillage, and design). 2. Emoji Mapping Dataset: This includes 392 common emojis mapped to their Persian equivalents to enhance sentiment analysis accuracy. 3. Sentiment Classification Results: This section includes three CSV files that detail sentiment classifications biased toward specific perfume attributes: Scent, Longevity, and Sillage. These classifications are derived from user comments using the ParsBert model, integrating user ratings to provide a nuanced understanding of consumer preferences. For access to the dataset and further details, please visit our GitHub repository. Total number of user comments: 36,434 Total number of unique users: 7,387 Number of emojis in mapping: 392 Number of CSV files with sentiment classifications: 3