Predicting Analysis of the Multi-Sensor Signals in Terms of Time Series

This study is aimed to present a new way to predict multi-sensor signals values based on the given values registered before. The method is based on the generalised serialisation method, proposed by the author earlier, and autoregression with adjustable window width. The given signals have been received as a sequence of values related to a random instances of measurements. However there exist some patterns that imply the values to be not random in fact. It is natural in this proposition to consider all the signals came from the object to be predictable up to some precision due to the pattern for every signal exists. It is shown in the paper how to make such unification of the signals to receive the time series; there are various methods of prediction for the time series, and autoregression is chosen for this study; the generalised serialisation method allows to transform the value predicted by autoregression to the multi-values needed for prediction.