TALENTED: An Advanced Guarantee Public Order Tool for Urban Inspectors

Abstract

In the streets of Chinese cities, we often see that illegal pedlars sell some fake and inferior products such as outdated food and inferior household goods to people who do not know about this, which may cause serious health problem. Besides, pedlars often cause people to gather and so may lead to traffic accidents. Thus, there are great requirements how to control illegal pedlars, and how to analyze, model and predict illegal pedlars activities. Such research will help urban inspectors decide better strategies to guarantee public order. Thus, in this paper, we explore this problem, and propose a model called TALENTED (Target Attributes LEarNing model with TEmporal Dependence) to deal with the problem. TALENTED provides three main contributions. First, a new learning model is proposed to predict the probability of each target being attacked, and our model consists of three aspects: (i) This model considers a richer set of domain features; (ii) Adversaries’ previous behaviors affect their new actions; (iii) Each target has different attributes and the adversaries weight them differently. Second, we adopt a game-theoretic algorithm to compute the defender’s optimal strategy. Finally, simulation results illustrate the reasonability and validity of our new model.