A method and an apparatus of identifying a user feature include: in response to receiving a designated evaluation of an interacting party in a service interaction, determining a reliability value of the designated evaluation of the interacting party based on a feature value of the service interaction; based on reliability values of multiple designated evaluations of the interacting party in multiple service interactions including the designated evaluation, determining a reliability mean value of the multiple designated evaluations; determining a reliability weighted mean value of the multiple designated evaluations based on a frequency weight corresponding to the multiple designated evaluations and the reliability mean value, wherein a number of the multiple designated evaluations is in a monotonically incremental or decremental relationship with the corresponding frequency weight; determining a user feature of the interacting party based on a size relationship between the reliability weighted mean value and a preset reliability threshold. Using the solution provided in the embodiment of the present disclosure can improve the accuracy of determining a user feature of an interacting party in a service interaction.