Predicting Mobile Usage and Call Detail Record Features from Personality Traits for Network Management Applications

Personality-informed predictions offer new opportunities for enhancing mobile network management by providing additional signals for applications such as resource allocation and proactive planning. In this paper, we explore how personality traits can be used to predict Mobile Internet Usage (MIU) and Call Detail Record (CDR) features, enabling administrators to initialize management tasks even when direct consumption history is unavailable. We collected data from ۶۷ participants, gathering MIU and CDR logs alongside personality traits obtained through questionnaires. From the usage logs, we extracted a diverse set of features, which were then modeled using Multiple Linear Regression (LR) and Gaussian Process Regression (GPR) to estimate demand and traffic variability. We further highlight the potential of these predicted features for managerial applications and then address their role in the cold-start scenario. The predicted features were subsequently incorporated into a Safe-Soft proportional fairness formulation of the bandwidth resource allocation problem, with an explicit comparison between allocations using personality-based priors and those relying on a uniform no-prior baseline. Evaluation results demonstrate that priors reduce demand shortfalls and user-level violations relative to the baseline, showing stable gains across repeated simulations and robustness to bandwidth and parameter variations. These improvements highlight the potential of psychology-informed priors as a complementary tool for cold-start allocation, supporting proactive planning and fairness in mobile network administration.

کلیدواژه ها:

personality traits ، Mobile Usage Data ، regression models ، Mobile Network Management ، Behavioral Data Analytics ، Cold-Start Problem