طريقة مقترحة لتحديد النماذج الكفوءة للسلاسل الزمنية

الملخص EN

The analysis of time series data is one of the most important statistical topics, usually focuses on forecasting the future behavior of the series at a certain time for certain purposes.

In general, there are sequence analysis steps which should be considered in a stationary time series to determine and provide a good statistical forecasting model, these steps are as follows: 1- Identification: Model selection According to the behavior studies of Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) for errors we could identify one of the following three types of time series forecasting models: • Autoregressive model AR (p) • Moving Average model MA(q) • Autoregressive-Moving Average model ARMA(p,q) Models fall under the heading of ARIMA models, and by using order selection criteria, then we could identify and determine the model's order p,q.

2- Estimation: Model parameters estimation 3- Diagnostic Checking 4- Forecasting A New Suggested Technique to Determine the Best Models: For any stationary time series, we take the ARMA(p,q) models with all its possible orders; 0, 1, and 2 for (p,q).

And for each model we check (ACF) and (PACF) for its series of estimated errors {ât}, then the model which will satisfy the equation (7) either (9) or both is chosen to be the best model.

This new suggested technique has been applied on actual 14 series data (Non-stationary series has been transferred to stationary series) which represents the raining quantities fallen on Darnah and Shahat zones in Libya within raining months for 58 years.

Research is attained to several conclusions and recommendations: 1.

In many cases of time series analysis, no one could be able to know the model's type through the behavior studies of (ACF) and (PACF), then the identification step is useless.

Our new suggested technique comes over this fatigued step.

2.

This new suggested technique focuses on all possible alternatives of ARMA(p,q) models.

The model which has ACF and PACF of {ât} non-significant with smallest AIC, BIC, ASBC, and SBC is chosen to be the best model.

3.

This new suggested technique is very simple technique and saves a lot of efforts and time to reach the best model.

4.

Since this new suggested technique is easy to recognize and understand , we strongly recommend to apply this new technique especially in education fields beside the ordinary time series analysis.