The main topic of my thesis is the introduction of ARIMA (autoregressive integrated moving average) time series model. In the first part I am describing the mathematical and statistical methods which are contained in ARIMA and are necessary to understand ARIMA time series model based on Box-Jenkins method.
Among other things, I am going to introduce a process that can resolve stationary problem when using stochastic time series model. In addition to that, I am going to describe the following algorithms: the linear regression, Gauss-Newton non-linear regression, Yule-Walker, Hannan-Rissanen and Maximum Likelihood Estimation. All of them are important in calculating the parameters of an ARIMA model.
In the second part of my thesis I am going to introduce an object-oriented program design based on the above mathematical-statistical methods. This can provide the structure of an ARIMA-based forecast program.
The last chapter contains the test results of the program I have implemented, comparing these results to the ones provided by a reference software also working with ARIMA time series model.