Proposed Loss Functions for Accurate Prediction of Terrorist Event Locations in Egypt | ||
| Journal of Contemporary Technology and Applied Engineering | ||
| Volume 3, Issue 2, March 2025, Pages 103-113 PDF (1.14 M) | ||
| Document Type: Original Article | ||
| DOI: 10.21608/jctae.2025.364093.1045 | ||
| Authors | ||
| Ghada Hamisa* 1; Loai Abdalslam2 | ||
| 1Electrical Engineering Dpt., Faculty of Engineering, Kaferelsheikh, Egypt. | ||
| 2Computer Science Developer, Alexandria, Egypt | ||
| Abstract | ||
| At the last decades, human security is threatened by terrorism. A rapidly growing field of study aims to understand terrorist attack patterns for counter-terrorism policies. Terrorist attacks can be analyzed and predicted with detailed historical data for better prevention and early warning. In this research, we use predicting geolocation in an open area space. It is aimed to predict the terrorist action location before it being occurred. Two novel ideas are presented in this research. The first idea is using spherical-distance measurements between two points instead of traditional straight-line distance measurements as previously used in predicting locations. Spherical-distance measurements depends on coordinate-form(x,y). The second proposed idea is coupling geolocation-functions to famous loss functions, like Mean Square Error (MSE) loss function to achieve accurate performance. The proposed geolocation-functions are “Haversine formula”, and “Equirectangular Projection formula”. The two proposed geolocation loss functions are hybrid to MSE loss function. The deep learning algorithm, Long Short-Term Memory (LSTM) is used for training our model. Experimental results show that the proposed loss functions achieved high accuracy compared to the traditional one. In this research, we use dataset of terrorist events in Egypt. | ||
| Keywords | ||
| Deep Learning; Long Short-Term Memory; loss function; Haversine formula; Terrorist events location prediction | ||
|
Statistics Article View: 380 PDF Download: 77 |
||
