Open Dataset for Predicting Pilgrim Activities for Crowd Management During Hajj Using Wearable Sensors

Abstract

This study aims to create and examine a multimodal dataset to enhance crowd management during the Hajj seasons. Sixty-four participants were engaged in Hajj rituals such as Tawaf, Saai, prayer, and Doaa providing location and peripheral physiological data, collected and annotated using a custom-made smartphone application. The collected data was leveraged to conduct a comprehensive analysis, specifically focusing on the classification of the type of Hajj activity, level of fatigue, and emotional states based on peripheral physiological signals. Three deep learning classification models were developed and validated using feedforward neural networks. The models achieved satisfactory accuracy scores in classifying the type of Hajj activity (41.71%), level of fatigue (85.27%), and emotional states (82.47%). While presenting a straightforward use case, this research chiefly provides decision makers and the scientific community with a statistically significant open data set aside with a deep learning architecture capable of characterizing crowd behavior for the purpose of automating crowd management and monitoring.