Have library access?
  • Journals

A Hybrid Arima Elman Artificial Neural Networks Approach With Overlapped Moving Window - An Experiment Study For Crime Rate Predicting


Crimes forecasting is an important area in the field of criminology for the government and can work more proactively to analyze and prevent crime. Crime data time series consists of complex linear and non-linear patterns and are difficult to forecast. Currently, Autoregressive integrated moving average (ARIMA) and Elman artificial neural network (EANN) model are popular used to develop time series model of the linear and non-linear type respectively. Moreover, Overlapped moving-window is also applied to improve prediction accuracy. However, each method has its limitation and to investigate the effects of the overlapped moving-window method on predicting the trend of crime rates, hence, a hybrid model is proposed to take advantage of the feature and strength of ARIMA and EANN approaches trough an overlapped moving-window with the aim of improve forecasting performance. The novel model can provide relevant information for resource planning and decision making in time-series predictions of police. Furthermore, the comparison showed that the proposed model has optimum prediction results. Also, the hybrid method in this work has increased the efficiency of the forecast and managed to reduce the forecasting errors.