Abstract: Aspects automatically minimize marketing rates as a function of optimized response rate models. Response rates are estimated from marketing offers targeted from a first entity to topic-entity pairs at respective specified marketing rates as a function of relative differences in their consumer sentiment scores. Marketing offers are targeted to a subset of the topic-entity pairs that each have estimated response rates meeting a threshold response rate constraint. Respective costs and actual rates of response from consumers are determined for the targeted offers. Response rates are modeled to determine modeled response parameters for the topics and entities of the subset topic-entity pairs as regressions of minimized differences in value between estimated and actual response rates. Marketing rates of the subset entity-topic pairs are minimized to meet the threshold response rate constraint and a marketing cost constraint as a function of the modeled response parameters.

Abstract: Aspects automatically minimize marketing rates as a function of optimized response rate models. Response rates are estimated from marketing offers targeted from a first entity to topic-entity pairs at respective specified marketing rates as a function of relative differences in their consumer sentiment scores. Marketing offers are targeted to a subset of the topic-entity pairs that each have estimated response rates meeting a threshold response rate constraint. Respective costs and actual rates of response from consumers are determined for the targeted offers. Response rates are modeled to determine modeled response parameters for the topics and entities of the subset topic-entity pairs as regressions of minimized differences in value between estimated and actual response rates. Marketing rates of the subset entity-topic pairs are minimized to meet the threshold response rate constraint and a marketing cost constraint as a function of the modeled response parameters.

Abstract: A graph is constructing using a user node of a user and a set of connection nodes corresponding to connections in a social network of the user. The graph is overlaid on a risk map, which is a mapping of a geo-social risk characteristic of a geographical area. A first and a second risk indices are determined for a first and a second connection in the set of connections. Using the risk map, a first and a second risk diffusion boundary of the first and the second risk index are computed. When the first risk diffusion boundary includes the user node, a first risk contribution from the first connection node to the user node is computed using the first risk index. Using the first risk contribution, a risk index of the user is determined to accounts for a risk attributable to the user due to the user's social network.