The present application discloses a monitoring indicator optimization method of a business operation analysis early warning system. An early warning analysis method comprises three stages which are a threshold design phase, a static early warning stage and a dynamic early warning stage. The threshold design is based on large sample historical data, a Monte Carlo simulation method is used, and a threshold can be set according to the development change trend of an operational performance indicator. According to static early warning, a multi-layer radar chart analysis method is statically used, and the identification of a key indicator which generates unusual action is facilitated. According to dynamic early warning analysis, a prediction of combining logic regression and a neural network is used, and the indicator change trend can be predicted. Thus, according to the method, the corresponding reaction of the change trend of the operational performance indicator can be effectively carried out, and the change trend of the operational performance indicator can be quantitatively analyzed. In addition, through the method, the problems of insufficient unusual action identification and prediction functions of the key indicator and weak support ability of large sample data in the previous comprehensive early warning can be solved.