Machine Learning in Digital Retail - Demand Forecasting for Inventory Management in a Sportswear Company - Internship Report

Aug 4, 2026

Upload by Levi

Accurate demand forecasting supports better decisions across the retail value chain, especially through inventory management and safety stock planning. This internship project improves the safety stock process for Always Available (AA) products in a sportswear company’s digital channel by forecasting demand units using machine learning. Tree-based models—Random Forest, XGBoost, and LightGBM—are evaluated with MAE and RMSE, considering effects of promotions, inventory levels, website traffic, and conversion rates. Results show comparable MAE across models, while LightGBM achieves the lowest RMSE, indicating stronger robustness and slightly higher predictive performance.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages52

Reading Time2 h 11 min

AccessFree to download

1. Introduction

1.1. Context

1.2. Motivation and Objectives

2. Literature review

2.1. Introduction

2.2. Forecasting Methods in Retail

2.3. Factors Influencing Demand

2.4. Summary

3. Methodology

3.1. Business Understanding

3.2. Data Understanding

3.2.1. Data Collection

3.2.2. Data Analysis and Visualization

Master Degree Program in Data Science and Advanced Analytics MDSAA Machine Learning in Digital Retail Demand Forecasting for Inventory Management in a Sportswear Company Andreia da Silva Tavares Bastos Internship Report presented as partial requirement for obtaining a Master’s Degree in Data Science and Advanced Analytics NOVA Information Management School Instituto Superior de Estatística e Gestão de Informação Universidade Nova de Lisboa NOVA Information Management School Instituto Superior de Estatística e Gestão de Informação Universidade Nova de Lisboa Machine Learning in Digital Retail Demand Forecasting for Inventory Management in a Sportswear Company by Andreia da Silva Tavares Bastos Internship Report presented as partial requirement for obtaining the Master’s degree in Data Science and Advanced Analytics, with a specialization in Business Analytics Supervised by Mauro Castelli, PhD, Nova University May, 2024 STATEMENT OF INTEGRITY I hereby declare having conducted this academic work with integrity. I confirm that I have not used plagiarism or any form of undue use of information or falsification of results along the process leading to its elaboration. I further declare that I have fully acknowledged the Rules of Conduct and Code of Honor from the NOVA Information Management School. Portugal, 28-05-2024 DEDICATION This project is dedicated to my parents and brother, my inspirations in life. To my boyfriend as well, without whom I could not have succeeded. Your love and support meant the world to me. And last but not least, to my dear friends, who were always there for me. Thank you for everything. ACKNOWLEDGEMENTS I would like to thank my supervisor Mauro Castelli for the support and encouragement throughout this process. I also wish to extend my gratitude to my team, I couldn’t have done without you. ABSTRACT Several domains benefit from accurate forecasts, especially within the retail sector, as precise demand forecasting plays a key role in making informed decisions, such as for inventory management. The focus of this project is to improve the safety stock process for Always Available (AA) products in the digital channel in a Sportswear company, by implementing a machine learning solution to forecast demand units. By providing an accurate forecast, the project aims to optimize inventory levels, ensuring that sufficient stock is maintained to meet customer demand without overstocking. This approach focused on the implementation of tree-based models, specifically Random Forest, XGBoost, and LightGBM, which were evaluated using two performance metrics: Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) . Various factors influenced the models' performance, including promotions, inventory levels, website traffic, conversion rates, etc. All three algorithms demonstrated similar MAE scores, highlighting their effectiveness in handling complex and large datasets and showing that tree-based models can provide highly effective solutions in aretail context. However, LightGBM emerged as the most robust method, achieving the lowest RMSE and thereby indicating a slightly superior performance. This project demonstrates the significant benefit that machine learning can bring to the retail sector, driving business growth. KEYWORDS Digital Retail; Machine Learning; Tree-Based Regression Algorithms; Inventory Management. Sustainable Development Goals (SDG): TABLE OF CONTENTS 1. Introduction .................................................................................................................. 1 1.1. Context .................................................................................................................. 1 1.2. Motivation and Objectives .................................................................................... 1 2. Literature review ...............................................................................

FAQs

What is the project objective in the internship report?

The project aims to improve the safety stock process for Always Available (AA) products in the sportswear company’s digital channel by forecasting demand units.

Which machine learning models are used for demand forecasting?

The report implements tree-based models, specifically Random Forest, XGBoost, and LightGBM, to predict demand units.

How are the models evaluated and which model performs best?

Models are evaluated using MAE and RMSE, with MAE scores remaining similar across all three methods. LightGBM achieves the lowest RMSE, making it the most robust approach.

Aug 4, 2026

Upload by Levi