Study On Sparse Feature Selection Of Terrorist Attacks Based On Hard Threshold Function

As terrorist attacks occur frequently around the world,countries have also strengthened their efforts to combat terrorist attacks,but terrorists are still rampant.Therefore,research on the main factors affecting the occurrence of terrorist attacks has become increasingly important.This article focuses on the data of 2009-2018 in the global terrorism database,and mainly applies the sparse recovery algorithm to feature selection.For the variables in the data,a descriptive statistical analysis method is used.First calculate the Gini coefficient of weapon type,attack type,area and target or victim type;secondly analyze the casualties of the people in the past decade and the top ten countries that have been attacked by terrorism;and finally analyze the degree of property damage.For the selection of variables for terrorist attacks,the following two types of algorithms are mainly used: one is the PCA algorithm and the LASSO algorithm,and it is found that the claimed responsibility,target or victim subtype,weapon subtype,international ideology and the occurrence of terrorist attacks closely related.The other type proposes a sparse feature selection algorithm based on the sparse recovery algorithm.Numerical experimental results show that the sparse feature selection results of the OMP algorithm and the IHT algorithm are the same.According to the comparison of the running time and error of the two algorithms,the IHT algorithm is more suitable for the sparse feature selection of terrorist attacks.The sparse feature selection results of the four algorithms indicate that suspected terrorism,weapon subtype,international miscellaneous,international ideology,part of the incident group,nearby area,target or victim subtype are the main factors affecting the occurrence of terrorist attacks.From the results of sparse feature selection to analyze the incentives for terrorist attacks,in order to prevent or reduce the occurrence of such incidents.