Link Prediction Based on Sequential Bayesian Updating in a Terrorist Network

Link prediction techniques are being increasingly employed to detect covert networks, such as terrorist networks. The challenging problem we have been facing is to improve the performance and accuracy of link prediction methods. We develop an algorithm based on Sequential Bayesian Updating method that combines probabilistic reasoning techniques. This algorithm adopts a recursive way to estimate the statistical confidence of the results a prior and then regenerate observed graphs to make inferences. This novel idea can be efficiently adapt to small datasets in link prediction problems of various engineering applications and science researches. Our experiment with a terrorist network shows significant improvement in terms of prediction accuracy measured by mean average precision. This algorithm has also been integrated into an emergency decision support system (NBCDSS) to provide decision-makers’ auxiliary information.