YSocial - RecSys Simulations

Dataset Description This dataset contains the SQLite databases used to generate the empirical results reported in the paper. Each database file corresponds to one independent simulation run for a specific network topology and recommender. The dataset is organized as one compressed archive per recommender: icwsm2027_recsys_visibility_data/ RC.zip P.zip UCF.zip ICF.zip F.zip FP.zip LR.zip Each recommender archive contains two folders, one for each topology: <recommender>.zip BA/ <recommender>_BA_run_00.db ... <recommender>_BA_run_09.db ER/ <recommender>_ER_run_00.db ... <recommender>_ER_run_09.db For example, UCF.zip contains the user-user collaborative filtering runs. The file BA/UCF_BA_run_03.db is the fourth UCF run on the BA topology. Recommenders RC: reverse-chronological baselineP: popularity-based recommenderUCF: user-user collaborative filtering recommenderICF: item-item collaborative filtering recommenderF: social-feed baselineFP: friend-popularity recommenderLR: LR recommender, as defined in the paper Topologies BA: Barabasi-Albert topologyER: Erdos-Renyi topology For each recommender and topology, the dataset contains 10 independent simulation runs, indexed from run_00 to run_09. In total, the dataset contains 140 database files: 7 recommenders x 2 topologies x 10 runs = 140 runs The databases are intended to support reproduction of the aggregate analyses and figures reported in the paper. Users can extract the relevant recommender archive and read the files as SQLite databases.