Application of Artificial Intelligence for Sales Forecasting in Retail
MAGAZINE №2-3 (103) - 2021
AUTHORS
INYUTSINA V.S. - Data Analyst, OZON (Moscow, Russia)
CATEGORY Supply chain planning Information technologies in logistics and SCM Modern concepts and technologies in logistics and SCM
ABSTRACT
Nowadays, e-commerce became one of the most dynamic technology markets in Russia. It is gradually becoming an essential part of the national economy. This trend has intensified especially during the pandemic, when online shopping became an affordable tool of necessity. The growing demand has determined the basic principles of leadership on the e-commerce market, such as having a wide coverage of regions, fast and free delivery due to a developed logistics infrastructure and accurate sales forecasting. At the current stage of e-commerce development, participants have entered the level of competition through the use of the latest technologies, among which are: artificial intelligence, augmented reality and machine learning tools.
For the company in 2021, the operational efficiency is at the most priority. Marketplace will focus on launching new and developing current products and services, increasing the supplier base, and expanding the logistics infrastructure.
The main result of this study is an analysis of the state of the logistics infrastructure in an online retail company, including a generalization of the number of pick-up points and automated post offices in each region, as well as their throughput. In addition, the long-term short-term memory (LSTM) neural network will be used as the main solution to the problem. The final result is the selection of various regions in which the existing logistics infrastructure of the company does not agree with the predicted volume of future sales, and a method for calculating the demand for a pickup point required to cover the demand will be proposed.
Before starting the experiment, a basic analysis of business and logistics will be carried out to understand the current state of the infrastructure. This will be followed by an analysis of the current process for managing the network of pick-up points, with a further understanding of the models and data used. The importance of this part is due to one of the goals of this study - it offers the company an innovative approach to the points of issue of orders. In addition, a review of networks about the available neural structures will be carried out with a justification of the model, which will be announced in the experiment. Possible examples of neural network structures for studying are recurrent neural networks, LSTMs, multilayer perceptrons, and others.
The company's research will expand market coverage, reduce lost sales and overall lead times.
Keywords:
