Summary: | 碩士 === 輔仁大學 === 應用統計學研究所 === 89 === Abstract
This study is to establish a forecasting model for construction cost indices based on the characteristics of neural network and ARIMA. We try to find the best result for the purpose of providing a reference for budgeting and bidding public infrastructure of contractor and construction companies. Since the neural network model can only provide point but not confidence interval estimation. Therefore, a semi-parametric prediction method is proposed to solve the above-mentioned drawbacks. The prediction methods incorporated into the system consist of neural network model that estimates the trend, as well as an ARIMA forecasting of the residual series. It terms of the adaptability of ARIMA methodology, the forecasting intervals of the system can be successfully constructed.
Analytic results demonstrate that the designed neural network model provides a better forecasting result than the ARIMA model in terms of MSE, RMSE, MAE and MAPE. Besides, semi-parametric prediction method can successfully provides the asymptotic prediction intervals and hence provides an alternative approach for practitioners.
Key words:Construction Cost Indics,Neural Network,ARIMA
|