Intelligent Information System for Hajj Management by Using Data Mining Techniques

Abstract

seasons is a critical challenge due to the massive number of pilgrims and the limited capacity of the holy sites. Ineffective crowd management has historically contributed to severe congestion and life-threatening incidents. This study investigates the use of machine learning techniques to predict crowd density levels—Low, Medium, and High—using the publicly available Hajj and Umrah Crowd Management dataset. The dataset contains simulated records reflecting environmental conditions, temporal factors, health status, and pilgrim activities. Comprehensive preprocessing steps were performed, including handling missing values, scaling numerical variables, and encoding categorical attributes. Mutual Information–based feature selection was applied to identify the most influential predictors and reduce model complexity. Five classification algorithms were evaluated: Random Forest, K-Nearest Neighbors, Support Vector Machine, Decision Tree, and Logistic Regression. Model performance was assessed using both Hold-out (80–20 split) and 5-Fold Cross-Validation techniques