An improved algebraic forecasting method with internal smoothing is proposed for short-term time series prediction. The concept of the H-rank is proposed for the detection of a base fragment of the sequence. Numerical experiments with artificially generated and real-world time series are used to illustrate the forecast method.