SARIMA Method Research Articles

Forecasting production of fossil fuel sources in Turkey using a comparative regression and ARIMA model

This study aims at forecasting the most possible curve for domestic fossil fuel production of Turkey to help policy makers to develop policy implications for rapidly growing dependency problem on imported fossil fuels. The fossil fuel dependency problem is international in scope and context and Turkey is a typical example for emerging energy markets of the developing world. We developed a decision support system for forecasting fossil fuel production by applying a regression, ARIMA and SARIMA method to the historical data from 1950 to 2003 in a comparative manner. The method integrates each model by using some decision parameters related to goodness-of-fit and confidence interval, behavior of the curve, and reserves. Different forecasting models are proposed for different fossil fuel types. The best result is obtained for oil since the reserve classifications used it is much better defined them for the others. Our findings show that the fossil fuel production peak has already been reached; indicating the total fossil fuel production of the country will diminish and theoretically will end in 2038. However, production is expected to end in 2019 for hard coal, in 2024 for natural gas, in 2029 for oil and 2031 for asphaltite. The gap between the fossil fuel consumption and production is growing enormously and it reaches in 2030 to approximately twice of what it is in 2000.

Seasonal decomposition and forecasting of telecommunication data: A comparative case study

In this paper, forecasting models for the monthly outgoing telephone calls in a University Campus are presented. The data have been separated in the categories of international and national calls as well as calls to mobile phones. The total number of calls has also been analyzed. Three different methods, namely the Seasonal Decomposition, Exponential Smoothing Method and SARIMA Method, have been used. Forecasts with 95% confidence intervals were calculated for each method and compared with the actual data. The outcome of this work can be used to predict future demands for the telecommunications network of the University.