We discussed in this work some predictive methods for time series and it is decomposing time series to its component (trend, Seasonality, cycle, random), Exponential smoothing, ARIMA, then we discussed some combining methods, then we formed a new c
ombine for predict time series which depends on combining exponential smoothing and ARIMA using weighted average with MAPE weights, and applied all methods above on three seasonal time series , first hourly temperature in Aleppo in august 2011 ,second monthly milk production peer cow in Australia from Jan 1962 to Dec 1975,third quartly electricity production in Australia from Mar 1956 to Sep 1994, and compared the results which approved that the suggested method is the best.
