---

This paper studies an M/G/1 repairable queueing system with multiple vacations and N-policy, in which the service station is subject to occasional random breakdowns. When the service station breaks down, it is repaired by a repair facility. Moreover, the repair facility may fail during the repair period of the service station. The failed repair facility resumes repair after completion of its replacement. Under these assumptions, applying a simple method, the probability that the service station is broken, the rate of occurrence of breakdowns of the service station, the probability that the repair facility is being replaced and the rate of occurrence of failures of the repair facility along with other performance measures are obtained. Following the construction of the long-run expected cost function per unit time, the direct search method is implemented for determining the optimum threshold N* that minimises the cost function.

---

Three combination methods commonly used in tourism forecasting are the simple average method, the variance-covariance method and the discounted MSFE method. These methods assign the different weights that can not change at each time point to each individual forecasting model. In this study, we introduce the IOWGA operator combination method which can overcome the defect of previous three combination methods into tourism forecasting. Moreover, we further investigate the performance of the four combination methods through the theoretical evaluation and the forecasting evaluation. The results of the theoretical evaluation show that the IOWGA operator combination method obtains extremely well performance and outperforms the other forecast combination methods. Furthermore, the IOWGA operator combination method can be of well forecast performance and performs almost the same to the variance-covariance combination method for the forecasting evaluation. The IOWGA operator combination method mainly reflects the maximization of improving forecasting accuracy and the variance-covariance combination method mainly reflects the decrease of the forecast error. For future research, it may be worthwhile introducing and examining other new combination methods that may improve forecasting accuracy or employing other techniques to control the time for updating the weights in combined forecasts.