Selection of Basic Function for Identification of Time Series Based on Associative Majority Approach

Abstract

The paper proposes a model for choosing basic functions for automated identification of time series, an algorithm and software for identifying time series based on an associative majority approach, which allow to identify the type of basic dependencies of dynamic processes. The task of identifying the type of basic functions is solved using the classical theory of pattern recognition. Identification is performed by comparing the original image with the standards stored in a single image space. The image space is the area of device memory (associative memory) on which identification is performed. A feature of the proposed model is the efficiency of comparing the original images with the images of standards by using a single attribute space and the possibility of comparing all images in one measure. To increase the speed of identification in the algorithm for selecting basic functions, an associative majority approach to storing and searching identification data in electronic memory is also used. The proposed algorithm and software for selecting basic functions are universal, since they allow identifying the type of basic dependence in any dynamic process, regardless of the specifics of the subject area under study. Automated identification of the type of basic functions reduces the time to build predictive models and allows you to quickly predict further variants of the process flow.