Selection of time series forecasting model, using a combination of linguistic and numerical criteria

One of the problems in forecasting is selection a subset of suitable models, which perform accurate results not only in tested part of time series (TS), but in real forecast. In previous our paper a TS forecasting technique was proposed, where a framework of TS model selection using linguistic and numerical criteria was proposed. To choose a subset of suitable models from a given set of TS models we developed a linguistic description of TS behavior based on identification of TS general tendency. In this paper we focus on extension of a linguistic description of TS behavior and study it's efficiency in out of sample TS part. The study of a proposed framework in forecasting of 91 TS showed the improvement of accuracy in comparison with TS model selection, which used a numerical criterion only.