Framework Methodology for Creating Software Prototypes Focused On Machine Learning Applications in Accounting

The aim of the article is to present an original view on the methodology of software development using machine learning mechanisms in the problems of modern accounting.  In particular, this applies to the possibility of using existing base methodologies such as Scrum, UP (and its derivatives), CRISP-DM. The reasons for the great interest in agile methodologies were systematized and characterized. The basic methodologies were characterized and attention was drawn to the problems with their use in their pure form in the development process
of machine learning-oriented software. Proposed and described is an original methodology for creating a software prototype for automating the processing of banking operations in order  to map transactions with invoices. The conclusion contains a synthesis of the presented approach and an indication of directions for further research in the considered area.