A Review of Predicting Illegal Gatherings Using Machine Learning Methods

Nowadays, social networks have become a governing space and a suitable platform for the enemies of the country. During occasional protests that occur in the country, it has been properly demonstrated that social networks and foreign messengers serve as a place for organizing, managing, inciting, encouraging, and even educating young people for disruption and vandalism. The aim of this study is to show how machine learning can be utilized by security or law enforcement agencies to detect, prevent, and counter illegal gatherings with high precision and speed. To achieve this goal, a total of 73 articles published between 2012 and 2023, which employed machine learning methods, were examined. However, artificial neural networks were the most common method used, accounting for 44%, followed by random forest methods at 30%, and K-nearest neighbor methods at 26%. Additionally, 62% of researchers utilized online criminal datasets from public portals on the internet, while 38% used official and private datasets from legal organizations such as the police in their research. The results indicate that the use of random forest methods had the best performance, but for large datasets, the use of artificial neural networks yielded the best results for predicting crime occurrence based on time and location.