Crime Modeling with Truncated Lévy Flights and the Effects of Law Enforcement Actions
C. Pan,Bo Li,Yuqi Zhang,N. Geldner,Chuntian Wang,Li Wang,A. Bertozzi
2016-01-01
Abstract:In this paper, we develop a statistical model of criminal behavior with the application of the truncated Lévy flights, motivated by the previous crime models applying the random walks as in [M. B. Short, M. R. D’Orsogna, V. B. Pasour, G. E. Tita, P. J. Brantingham, A. L. Bertozzi, and L. R. Chayes, Math. Models Methods Appl. Sci., 18 (2008), pp. 1249-1267] and the Lévy flights as in [S. Chaturapruek, J. Breslau, D. Yazdi, T. Kolokolnikov, and S. G. Mccalla, SIAM J. Appl. Math., 73(4) (2013), pp. 1703–1720]. On the discrete level, we assume that each criminal travels through space independently according to a truncated Lévy flight (that is, with a limited jump range reflecting a limited traveling distance) biased towards the attractive burglary sites characterized by a dynamic attractiveness field. Thus, the models in the previous works mentioned above can be treated as special cases of our model, with the truncation size varying from a single step size to the infinity. Then based on this lattice model, we derive a continuum limit, which is in good agreement with the discrete system. We observe that a local operator arises rather than a non-local operator in the continuum system, essentially due to the fact that the generator of the truncated Lévy process is the Laplace operator. Next we incorporate the law enforcement agents into the model following the ideas in [P. A. Jones, P. J. Brantingham, and L. R. Chayes, Math. Models Methods Appl. Sci., 20 (2010), pp. 1397-1423]. To compare the effects of different patrolling strategies (including an unbiased random walk, a biased random walk, and a truncated Lévy flight) quantitatively, we evaluate the total number of burglary events occurring in a given time frame and find that the truncated Lévy flight is the most effective in terms of reducing this quantity.