A hybrid assault detection system using random forest enabled xgboost-lightgbm technique

Abstract

This article presents the development of an assault identification system using face recognition in a closed location, by employing a  machine learning-based computer vision approach. The proposed model combines algorithms such as Random forest, XGBoost and  LightGBM techniques. The objective is to accurately identify and classify instances of assault in real-time based on facial recognition. The proposed approach utilizes machine learning algorithms to analyze facial features and patterns associated with assault activities. By  leveraging on a hybrid model, the system can be integrated into closed locations such as schools, workplaces, or public venues to  enhance security measures and promptly respond to potential threats. The findings of this research contribute to the field of computer vision-based assault identification systems, in addressing security challenges. Further advancements of the proposed hybrid model can  lead higher performance levels in various real-world scenarios and enhancing public safety and security. The system's performance was  evaluated using various metrics, including precision, recall, F1 score, accuracy, and ROC score. The results shows that the proposed system outperformed the existing system with its identified weakness and limitations of: Limited Robustness in Handling Complex  Variations, Inability to Handle High-Dimensional Data, Limited Scalability, e.tc. The hybrid model achieved impressive results, with a  precision of 98%, recall of 98%, F1 score of 97.7%, accuracy of 97.5%, and ROC score of 97.4%. The above findings demonstrated the effectiveness and robustness of the developed system in accurately detecting and recognizing assault instances within a closed location.