عنوان مقاله [English]
نویسندگان [English]چکیده [English]
The time series is to try it with the last of the series, a possible model for generating data to identify and comment on this model series is about the future behavior. Possible models for time series of the same patterns are random processes. General trend in the time series is quite variable and not predictable, is uncertain. Random by random changes in the level and slope processes that occur in the series are identified. Generally, quantitative methods for time series forecasting methods can be divided into variable and multivariate methods. Data used in this study using a quantitative variable is checked. Variable in a method of analyzing past data to identify an appropriate model and assuming changed model; it predicts the future behavior of the series are. Quantitative methods with a variable in a series of four methods that can be fitted model for the series include: extrapolation of the curve, smooth exponential, the Holth- winterz and Box -Jenkins. In the present series as an experimental station of the four methods were analyzed And the diffraction data set at the top station of the curve is the best way to analysis extrapolation model was selected. Note that in addition to the monthly data showed a seasonal behavior for a series of linear trend and seasonal factors,Decomposition procedures were used. Decomposition procedures for the analysis of the series using the least squares regression line fitted to the data found. Multiplicative model in the next step in the process of sharing data, trend data are lacking. Then central moving average (with a length equal to the length of season), no trend data were mooted In each period, the median seasonal values of raw seasonal adjustment were calculated and then finally, the seasonal index for the data to the data was seasonally adjusted. Final results of time-series model to estimate changes in the scale of seasonal dust extraction for the stations studied, with graphs and charts to predict the final residual normal probability model was outlined. The remainder of the normal probability plots of the model cannot be assumed to be normal for the remaining stations of the on the occasion of his reason, and this model is fitted to time series data.