Forecasting time series using neural networks on the example of primary sales of a pharmaceutical company

Abstract

This article is devoted to neural network forecasting of primary sales of a pharmaceutical company. The research includes an analysis of the main forecasting methods, various types of neural network architectures used to predict time series, as well as an analysis of neural network training methods. The primary sales of the pharmaceutical company AKRIKHIN JSC were considered as the applied basis of the study. As part of the study, an analysis of the revenue of the pharmaceutical company AKRIKHIN JSC from 2016 to 2021 was carried out, the selection of the main hyperparameters of the neural network was carried out and a forecast of primary sales for 4 weeks ahead was obtained. The practical significance of the research results is due to the fact that forecasting primary sales gives management an understanding of which direction the company is moving in and whether additional management decisions are needed to meet the planned values. This study can be used as an alternative method of sales forecasting, so the results will be in demand among analysts of pharmaceutical companies in sales planning departments.