Data mining of small sample times series on detection non-stationarity properties

In this paper, a Data Mining of real times series data using Self-correlation Coefficients is investigated. The standard KPSS method was performed on real Time Series Data, in an attempt to experimentally investigate and establish the connection between small Time Series data and hidden information relating to the properties of non-stationary Time Series. The aim of this paper is to resolve this problem using a Kernel practice algorithm which is based on the extraction of Self Correlation Coefficients (SCC) combining with a non-parametric spectral density approach. In this way, the SCC is applied as suitable filter which transforms the spectral coefficients of the FFT non-parametric procedure in a parametric one such as AR. The proposed method has yielded correct classification scores in the range of versus to AR model in the small samples.