GHENABZIA Ahmed
غنابزية أحمد
ahmed.ghenabzia@univ-msila.dz
0557100777
- MI - Joint Basci Teaching Department
- Faculty of Mathematics and Informatics
- Grade MAB
About Me
Location
Msila, Msila
Msila, ALGERIA
Code RFIDE- 1991-08-25 00:00:00
-
GHENABZIA Ahmed birthday
- 2025-08-31
-
2025-08-31
A Simulation-Based Behavioral Clustering Method for Crowd Dynamics Evacuation Analysis
Traffic management, urban planning, and emergency management cannot be efficiently done without crowd simulation. This paper proposes a Behavioral Clustering Method (BCM), which tackles the problem of forming crowds in clusters or subgroups based on fundamental behaviors so that congestion is minimized during effective evacuation processes. We designed BCM based on synthetic data obtained from the simulation of the evacuation of a crowd in high-risk situations. Our method regards pedestrians as intelligent agents and predicts key behavioral aspects of future crowd evacuations before they occur. We use cluster analysis on those movement and behavioral data for building as well as evacuation-friendly control strategies by clustering people into subgroups of behavioral similarity. The credibility of the model is validated through Python-based animations to detect and rectify errors. Results from simulation performance evaluations indicate that BCM is successful in modeling the evolution of crowd behavior at the time of evacuation.
Citation
Mohamed Chatra , ghenabzia ahmed , Samir Akhrouf , Mustapha Bourahla , , (2025-08-31), A Simulation-Based Behavioral Clustering Method for Crowd Dynamics Evacuation Analysis, HAUT, Vol:23, Issue:8, pages:20, Open Access
- 2025-05-22
-
2025-05-22
Urban Traffic Prediction Using Hybrid XGBoost–LSTM Model
This study introduces a novel hybrid predictive model that integrates eXtreme Gradient Boosting (XGBoost) with Long Short-Term Memory (LSTM) networks, specifically designed for real-time urban traffic congestion prediction. The proposed model innovatively incorporates external data, such as weather conditions, traffic incidents, and road classifications, and effectively addresses the common issue of class imbalance in the traffic dataset and captures dynamic spatiotemporal traffic relationships. This comprehensive approach enables the capture of complex, dynamic spatiotemporal traffic relationships more accurately. XGBoost performs robust feature selection and preliminary classification, generating probabilistic traffic jam level estimates. These outputs are subsequently enhanced through bidirectional LSTM layers that leverage temporal dependencies within traffic data, thus significantly improving predictive accuracy. The hybrid XGBoost–LSTM model was evaluated using approximately three million real-time traffic records from central London, providing a substantial and realistic testing environment. The results demonstrated its superior performance, achieving an accuracy of 93%, with precision values between 86% and 96%, and recall between 84% and 97% across varying congestion scenarios, from free flow to heavy congestion. Notably, the inclusion of probabilistic feature augmentation successfully mitigated the impact of class imbalance, further enhancing reliability. Comparative analyses against traditional and standalone methods highlighted the proposed hybrid model’s substantial improvement in accurately differentiating traffic jam levels, making it a valuable tool for intelligent transportation systems (ITS). This research contributes significantly to urban traffic management strategies, supporting smoother traffic flow and congestion reduction.
Citation
Messaoud BECHERE , Abdelbasset Barkat , ghenabzia ahmed , Derya Yiltas-Kaplan, , (2025-05-22), Urban Traffic Prediction Using Hybrid XGBoost–LSTM Model, International Journal of Computing and Digital Systems, Vol:18, Issue:1, pages:1-15, scopus