Research Of Terrorist Prediction Algorithm Based On Behavior Change

The existing terrorism prediction models mainly predict the organizations’ future performance based on its’ past relations between context and behaviors, without the consideration of the changes of behavior attributes caused by the changes of organizations’ context attributes. The terrorism behavior prediction algorithm based on Culture Modeling(CM) constructs the behavior prediction model of organizations according to the relations between organizations’ context and behaviors. However organizations generally have the counter-investigation ability, thus the time, places and the behavior intensity will be changed. Most existing models do not consider the changes of the organizations’ context and the resulting behavior changes. Only Change Analysis and Prediction Engine(CAPE) model considers the continuous changes of organization behaviors. The basic idea is that the changes of organization behaviors may be caused by the changes of context. The model constructs the changing rules between context and behaviors by building a change table and analyzing the condition of organization changing their behavior. But when a context change vector does not meet the condition of any change rules, CAPE cannot predict a change of behavior. At this point, the model combines Sit CAST and CONVEX method to predict the organizations’ behaviors. But Sit CAST+CONVEX method has an exponential time complexity and a low prediction accuracy. In addition, the non-correlative, weak correlative and redundant attributes in the terrorist dataset also severely affect the algorithm prediction results of terrorism behaviors.Therefore, in order to use any context changes to predict organization behaviors and to unify the prediction process, considering the characteristics of the high dimensions and small samples of terrorism data, a terrorism prediction algorithm based on behavior changes and Bayes method is proposed. In order to effectively extract context features associated with behavioral changes, the changing relationship between context and behaviors is used to improve Spectral Cluster Based on Attributes’ Association(SCBAA). The main contents of this paper can be summarized as follows:(1) Considering the characteristics of the high dimensions and small samples of terrorism data, this paper proposes a terrorism prediction algorithm based on improved change table using Bayes method to predict organization behaviorsaccording to any behavior changes. It predicts organization behaviors on the change table due to the fact that Bayes method classifies high dimensions and small sample in a fast and efficient way. Thus, it improves prediction precision and computing efficiency. In addition, considering the continuing effect of the changes of the organization’s context on its behaviors, the weighted Bayes method with different time lags is used to predict the behavior of the organization. Experiments on multiple organization data of MAROB show that, the proposed algorithm is better than CAPE algorithm on accuracy and time complexity. Comparing with CAPE, the accuracy rate of the proposed algorithm is increased by 10%~15% and the running time is dropped by two orders of magnitude.(2) By using the changing relationship between context and behaviors, a feature selection algorithm based on behavior changes and spectral cluster(FSBC) has been proposed on the basis of SCBAA algorithm. The proposed method improves the ability to predict changes in behavior by only extracting the context features related to behavior changes, so as to reduce the effect of contexts that are independent of behavior changes. Experiments on 9 terrorist actions data in MAROB show that, the improved feature selection method based on behavior changes and spectral cluster perform better than the selected traditional feature selection methods on multiple weighted bayes model.(3) Based on the idea of modularization, a prototype system is designed and implemented, which includes feature selection based on behavior changes and spectral cluster algorithm, and multiple weighted bayes model for predicting terrorist behaviors.