Tools for Image Recognition as a Basis for Effective Crisis Management

Abstract

In the context of the instability of the development of Ukraine's economy, the problem of crisis phenomena emerging at enterprises is becoming particularly important today. Such a phenomenon is natural for any business entity and tends to periodically recur throughout its entire existence—from inception to liquidation. The practice of enterprise operation shows that, of all the properties inherent in a crisis state, the most characteristic is its suddenness. This fact is crucial in determining the essence of crisis management, which consists of recognizing weak signals that carry information about an impending crisis, with the goal of timely action to avoid it. The need for early detection (recognition) of a crisis situation at an enterprise, preventing its further development, and practical diagnostics necessitates the search for and development of effective crisis management methods. One possible approach to solving this problem is the application of situation recognition methods built upon pattern recognition theory and its methods. Currently existing pattern recognition methods are quite diverse, but they traditionally involve an image. In this work, the "pattern" (образ) is proposed to be understood as a situation—a structured description of a studied object or phenomenon, demonstrated as a feature vector. Each element of this vector is a numerical value of one of the features. The advantage of this approach is not only the possibility of solving the problem of an object's belonging to a certain class but also the development of classifying algorithms whose complexity is aligned with the length of the sample based on which the recognition is performed. This can be achieved by training a recognition system on training samples. An approach to solving the problem of enterprise crisis state diagnostics based on situation recognition was also proposed, though such a procedure was built upon classical methods of multivariate analysis.