Mining And Forecasting Of Terrorism

As the forces of global terrorist organizations continue to grow,much of the world is shrouded in the shadow of terrorism. Terrorist organizations have been attacking the city in a blatant and indiscriminate way, seriously affecting the normal life and public security of mankind.Their cruelty has been disgraced by all peace-loving people all over the world. The fight against terrorism has become an arduous and imminent task.However, there are relatively few studies on data mining to predict the behavior of terrorist organizations, and the existing prediction algorithms have low accuracy or too high time complexity to be applied.Based on the database of global terrorist organizations GTD(University of Maryland researchers carefully collected, choreographed each terrorist attacks from 1970 to 2014), This paper mainly do research on the terrorist organization database of missing behavior data and terrorist organization behavior prediction problem. the main work is divided into the following three parts:First of all, it combs the research on the mining and forecasting of terrorist organization behavior. This paper summarizes the application of data mining in counter-terrorism, the anti-terrorism process based on data mining, expatiates on the existing algorithms of terrorist organization behavior prediction, and introduces the concept of association rules and the commonly used terrorist attacks Database briefly.Secondly, A hybrid classifier based on missing data of weapon data in the database of terrorist attacks is constructed, and the hybrid model is based on Naive Bayes, KNN and ID3. Experiments show that the mutual information feature is five times as random in the accuracy of model by comparing the mutual information feature with the stochastic feature. Model fusion also has higher prediction accuracy than single model, and the F1 value of weapon information prediction is as high as 0.645.Thirdly, A weighted rule rule mining model of temporal logic behavior is constructed and a LBT_Weight algorithm is proposed. The model considers the problem of terrorist organization behavior prediction as NP problem, and uses association rule algorithm to mine the frequent pattern of terrorist organization behavior. A series of contrast experiments are conducted to obtain a best model with a sliding time window of 7 days and a sliding step of 1 day with an exponential function as the attenuation function with attenuation factor a = 0.22.Compared with the existing Convexk_NN. algorithm, the LBT_Weight algorithm is 1.48% more accurate than Convexk-NN.