Modeling the forecast of temporal metrics based on method of reverse propagation of errors and optimization heuristics

This article discusses the problem of modeling the time series forecast in the field of mobile development, where you need to have data at what point in time users leave the application to take measures to increase the lifetime and income level brought by the application. The proposed algorithm based on a neural network trained by the back propagation method provides an opportunity to predict different time periods in relation to the number of remaining users in the application. The back propagation method is used in conjunction with heuristics to accelerate convergence and avoid hitting local minima. The scientific novelty is represented by the time series forecasting algorithm, which allows to increase the efficiency of decision-making procedures in intelligent analytical systems. The results of the forecast are presented for different time intervals, characteristic of the mobile application’s use.