Using machine learning to determine accounting codes based on the economic meaning of procurement documentation

Abstract

Large industrial enterprises in the daily processes of their activities generate, process, and store a huge amount of documentation, including procurement, which is required when interacting with various suppliers of necessary goods and services. This article describes an approach to solving the problem of automating the process of determining accounting codes based on the economic meaning of procurement documents using machine learning. The real data collected in the economic department of a trade industrial sector company were used in the volume of 1020 documents containing 183 different types of accounting code (class). The rationale of a quality metric for estimate developed models is given.

The study examined 14 different machine learning algorithms. The best result was shown by the Ridge Classifier algorithm, which showed an accuracy of 81% in determining the accounting codes.

The advantage of the above approach is a detailed description of the solution to the problem of determining the accounting code, based on the economic meaning of the text of the procurement documents.