In this diploma seminar thesis, we explore Singular Spectrum Analysis, a numerical method for analyzing and forecasting time series. The method is particularly useful for forecasting and decomposing time series into meaningful components, such as
trend, periodic components, and noise. We present the basic algorithm, analyze the selection of method parameters, and provide theoretical and practical guidelines for their determination. We find that there is considerable flexibility in parameter
selection, making it difficult to formalize this step of the method. We also describe some time series forecasting methods based on Singular Spectrum Analysis. The work includes several practical examples implemented in MATLAB.
|
