Efficient Algorithms for Time Series Prediction Method

This paper describes the effective algorithms for calculating the Krichevsky measure and the R measure which are used in the time series prediction method. This prediction method is based on universal coding and first was proposed in 1988. The reliability and potential of method’s application was investigated in the context of many real tasks, however, the prediction calculation speed was the main problem. The algorithms proposed in the work allow to calculate the predicted value online and have the minimum possible complexity which significantly expands the possibilities of applying the method itself.