Improving crime prediction through ensembles
Abstract
Predicting crimes is a challenge due to the complexity of the inherent temporal series in the phenomenon of interest, which often exhibit both linear and nonlinear components. Moreover, due to the dynamic nature of temporal patterns, depending solely on a single model can result in imprecise predictions. In this context, this study proposes the modeling and forecasting of crime temporal series using both individual models and ensembles, encompassing trainable and non-trainable approaches. Furthermore, a combination method is suggested involving the selection of a subset of individual models through a validation step. For the experiment, crime series from different locations, such as Pernambuco, Chicago, and Los Angeles, were employed. Based on the presented results, it is observed that ensemble models exhibited better predictive performance than the individual models used. However, regarding the proposed method, it is concluded that selecting a subset of individual models based on validation could be problematic when the validation sets are not sufficiently representative, whether due to a limited number of observations or distinct characteristics compared to the test set.
References
Calatayud, J., Jornet, M., and Mateu, J. (2023). Modeling noisy time-series data of crime with stochastic differential equations. Stochastic Environmental Research and Risk Assessment, 37(3):1053–1066.
Chen, P., Yuan, H., and Shu, X. (2008). Forecasting crime using the arima model. In 2008 fifth international conference on fuzzy systems and knowledge discovery, volume 5, pages 627–630. IEEE.
Cohen, L. E. (1981). Modeling crime trends: A criminal opportunity perspective. Journal of Research in Crime and Delinquency, 18(1):138–164.
Detotto, C. and Otranto, E. (2010). Does crime affect economic growth? Kyklos, 63(3):330–345.
Kelly, M. (2000). Inequality and crime. Review of economics and Statistics, 82(4):530–539.
Khairuddin, A. R., Alwee, R., and Haron, H. (2019). A review on applied statistical and artificial intelligence techniques in crime forecasting. In IOP conference series: materials science and engineering, volume 551, page 012030. IOP Publishing.
McDowall, D., Loftin, C., and Pate, M. (2012). Seasonal cycles in crime, and their variability. Journal of quantitative criminology, 28:389–410.
Olligschlaeger, A. M. (1997). Artificial neural networks and crime mapping. Crime mapping and crime prevention, 1:313.
Safat, W., Asghar, S., and Gillani, S. A. (2021). Empirical analysis for crime prediction and forecasting using machine learning and deep learning techniques. IEEE access, 9:70080–70094.
SILVA, E. G. d. et al. (2021). Uma abordagem de seleção dinâmica de preditores baseada nas janelas temporais mais recentes.
Sinha, B. B. and Biswas, T. (2023). An efficient framework for forecasting of crime trend using machine learning technique. In Proceedings of International Conference on Data Science and Applications: ICDSA 2022, Volume 2, pages 741–755. Springer.
Tariq, H., Hanif, M. K., Sarwar, M. U., Bari, S., Sarfraz, M. S., and Oskouei, R. J. (2021). Employing deep learning and time series analysis to tackle the accuracy and robustness of the forecasting problem. Security and Communication Networks, 2021:1–10.
Veiga, J. E. D. (2014). O âmago da sustentabilidade. estudos avançados, 28:7–23.
Zhang, G. P. (2007). A neural network ensemble method with jittered training data for time series forecasting. Information Sciences, 177(23):5329–5346.
Zhuang, Y., Almeida, M., Morabito, M., and Ding, W. (2017). Crime hot spot forecasting: A recurrent model with spatial and temporal information. In 2017 IEEE International Conference on Big Knowledge (ICBK), pages 143–150. IEEE.
