An algebraic segmentation method based on algebraic predictor with internal smoothing is proposed. The concept of the rank of the sequence is proposed for the detection of a base fragment of the sequence. Numerical experiments with a real-world financial time series illustrate the segmentation method.