This paper contributes to the field of dynamic social networks by developing a new methodology applicable to assessing a particular node's actual influence. Sensitivity analysis applied to the trained neural network further aids in explaining other nodes' involvement in the entire process.