Grey Systems: Theory and Application

Purpose – The purpose of this paper is to propose a model for effective data filling and precise prediction, which is used to solve the prediction problem of sequential data with the characteristics of poor information, high growth and containing extraordinary points.

Design/methodology/approach – After proving that the three principles of smooth sequence are not a sufficient condition for the judgement of sequence smoothness, judgement rules for sequence smoothness based on smoothness efficiency is introduced. Based on the non-homogenous discrete grey model (NDGM) model which fits for high growth sequence, model error caused by equal weight mean value is analyzed, and mean value generation weight efficiency is optimized by the method of differential. Prediction steps that fit sequences with high growth, poor information and containing extraordinary points is established on the basis of equal weight mean value generation efficiency.

Findings – The results are convincing: previous judgement rules used for sequence smoothness do not fit for the high growth sequence, new judgement rules introduced are more effective for high growth sequence. Sequence filling algorithm based on differential ration not only improve the filling of high growth sequence, but also enhance the prediction precision of these sequences.

Practical implications – The method exposed in the paper can be used to solve the prediction problem of sequences with poor information, high growth and containing extraordinary points, and it was proved in the cases of large and medium company new products income and Ufida Software Company. What is more, the method is also helpful in aspects of corporate financial control and strategy-making process.

Originality/value – The paper succeeds in proposing a new interpolation algorithm that is superior to ordinary mean value generation method in the aspects of generation and prediction and to grey interpolation algorithm in the aspect of information volume by defining sequence smoothness efficiency and introducing smoothness judgement rules that are easy to compute and fits for high growth sequence and not limited to monotonicity sequence.