Results obtained for the evaluation of the baselines and the proposed model for users.

We show results regarding Precision (P), Recall (R), Mean Average Precision (MAP), and pro-recovery suggested ratio (PRSR), neutral suggested ratio (NSR), harmful suggested ratio (HSR) and harmless suggested ratio (HLSR) of accounts at K accounts suggested. We also report the ratio of followed pro-recovery (PRFRS), neutral (NFRS), harmful (HFRS) and harmless (HLFRS) accounts over the number of accounts suggested of each type at K. We also calculate the ratio of followed pro-recovery (PRFRK), harmless (HLFRK) and harmful (HFRK) accounts over the total number of accounts suggested (K), along with the Average Precision-Harmlessness Ratio Score (APHR). In cursive, we highlight the best results for each combination of model and K.