Latent class model for analyzing response intervals between users on business chat applications

Abstract

These days many companies try to use communication data on business chat systems for the management of human resources. Especially, the reply time between a talker and a receiver is affected by the relationship with the receiver desired to establish. Therefore, the analysis model about the reply time between two employees as the minimum unit with using a chat system should be effective. In this research, we propose a latent class model that quantitatively expresses the relationship between the talker, the receiver and the reply time and analyze the communication characteristics from the viewpoint of the reply time. We can interpret the combination of high-dimensional data and the difference between user behavior by introducing multiple latent variables that are probabilistically occurred behind the observed variables. Finally, we apply the model to the real data of a Japanese company and show the usefulness of the model.