The algorithm of recognition over the moving average for prediction of dynamic series (Q2730559)

Let \(y_1,y_2,\ldots,y_{n}\) be a dynamic series. Using the formula of moving average NEWLINE\[NEWLINE\bar y_{t}^{m}= \bar y_{t-1}+(y_{t+p}- y_{t-(p+1)})/(2p+1),\quad \text{where} m=2p+1,NEWLINE\]NEWLINE the dynamic series is smoothed by polygonal lines \(y_{t}^{m}\), where \(m\) is a moving parameter. These polygonal lines can be used for prediction of the behaviour of the dynamic series. The authors propose an algorithm of recognition of critical points of the dynamic series which uses some set of moving polygonal lines and methods of pattern recognition.