comparison of auto regressive integrated moving average and artificial neural networks forecasting in mortality of breast cancer

Background & aim: one of the common used models in time series is auto regressive integrated moving average (arima) model. arima will do modeling only linearly. artificial neural networks (ann) are modern methods that be used for time series forecasting. these models can identify non-linear relationships among data. the breast cancer has the most mortality of cancers among women. the aim of this study was fitting the both arima and anns models on the breast cancer mortality and comparing the accuracy of those in parameter estimating and forecasting. methods & materials: we used the mortality of breast cancer data for comparing two models. the data are the number of deaths caused by breast cancer in 105 months in kerman province. each of arima and anns models is fitted and chose the best one of each method separately, with some diagnostic criteria. then, the performance of them is compared a minimum of mean squared error and mean absolute error. results: this comparison shows that the performance of anns models in parameter estimating and forecasting is better than arima model. conclusion: it seems that the breast cancer mortality has a non-linear pattern, and the anns approach can be more useful and more accurate than arima method.