Authors:
Shanmugapriya Murugavel, Francisco Hernandez

Addresses:
1,2Department of Mathematics, Munster Technological University, Cork, Munster, Ireland. s.murugavel@mycit.ie1, francisco.hernandez@mtu.ie2 

Abstract:

Forecasting techniques are widely used in various sectors, from predicting climatic changes, healthcare systems, agriculture crop simulation, and stock market management to business forecasting, including demand, supply, and sales forecasting. Sales prediction is one of the forecasting methods that predict the retailing of products to be sold in the future. Sales forecast additionally serves as a tool for identifying benchmarks, determining incremental results of new initiatives, planning resources to meet demand, and projecting future costs. The forecasting process helps retailers stock products based on customer requirements, improving the supply chain management process. During unexpected situations like the pandemic, sales of a few items, such as milk, bread and toilet paper, were unpredictable. This is an example where sales prediction plays a major impact. This paper compares the performance of various machine learning and statistical time series models in predicting sales. As strong seasonal fluctuations were observed, several regression approaches like LSTM, Linear Regression and Random Forests were used to predict future sales using the historically available data. Finally, the models were compared based on the RMSE scores to evaluate the performances.

Keywords: Statistical and Machine Learning Methods; Forecasting Retail Sales; Linear Regression; Supply Chain Management Process; Sales Forecast; Predicting Climatic Changes, Healthcare Systems.

Received on: 30/12/2022, Revised on: 27/02/2023, Accepted on: 07/04/2023, Published on: 17/04/2023

FMDB Transactions on Sustainable Computer Letters, 2023 Vol. 1 No. 2, Pages: 76-102

• Views : 160
• Downloads : 10

Download PDF