IIoT Based Trustworthy Demographic Dynamics Tracking With Advanced Bayesian Learning

Abstract

Tracking demographic dynamics for the built environment is important for a smart city. As a kind of ubiquitous Industrial Internet of Things (IIoT) device, portable devices (e.g., mobile phones) afford a great potential to achieve this goal. Tracking the demographic dynamics illuminates two things: populations mobility (where do people go) and the related demographics (who are they). Many past studies have investigated the tracking of population dynamics; however, few of them tried tracking the demographic dynamics. In this context, our study proposed a ubiquitous IIoT based trustworthy approach for built environment demographic dynamics tracking. First, we employed a meta-graph-based data structure to represent users life patterns and projected them into a low-dimension space as uniform features. Then, based on the life-pattern features, we derived a variation-inference-based advanced Bayesian model to infer the demographics. Finally, taking a region in Tokyo as a case study, we compared our methods with baseline methods (heuristic algorithm, deep learning), and the result proved a superior accuracy (the MAPE improved by 0.07 to 0.28) as well as reliability (0.78 Pearson correlation coefficient with survey data).