ARIMA model and user regulation-based dynamic variance real-time alarming method

The invention discloses an ARIMA model and user regulation-based dynamic variance real-time alarming method. The method comprises the following steps of performing longitudinal prediction to obtain prediction data X' of T data collection cycles, and recording real data X; performing transverse prediction to obtain a variance S ' of a corresponding time segment in a current cycle; performing linear combination to obtain a dynamic variance-based dynamic threshold according to weight values, endowed by a user, of all time segments and the variance S '; and when the value of //X-X'// is greater than the dynamic threshold, triggering an early warning apparatus. According to the method, a reliable prediction result is provided for a dynamic variance early warning mechanism by adopting an ARIMA time sequence prediction model, so that the early warning accuracy of the early warning mechanism is improved; the early warning error rate caused by normal data fluctuation due to environment change is reduced, namely, the e