Neural Network Based Hybrid Prediction Models for Healthcare Applications

Prediction models based on different concepts have been proposed in recent years. Improving the accuracy of prediction models has remained as a challenging task for researchers. The prediction accuracy depends not only on the model but also on the complexity of the data. Hence, it is important to choose the best model based on the complexity of data in the prediction. The time series prediction model determines future trends based on past values and corresponding errors. There are many models used for time series prediction, such as statistical techniques including linear regression, moving average, exponential smoothing, autoregressive moving average (ARMA) and autoregressive integrated moving average (ARIMA) and soft computing methods, such as artificial neural networks and adaptive neuro-fuzzy inference system. In the development of time series analysis, it is well known that many phenomena are nonlinear, namely the relationship between the past and current events is nonlinear. Thus, the linear time series models are not sufficient and appropriate for these cases. As a consequence, nonlinear time series models have gained importance in time series prediction. The accuracy rates obtained using linear time series models may not high as they have limitations in handling the non-linear relationships among the data. Models based on artificial neural networks model are considered to be better in handling such non-linear relationships. In the real-world, the time series data consist of complex linear and nonlinear patterns and it may be difficult to obtain high prediction accuracy rates using only linear or neural network model. Hybrid model which combines both linear and neural