Unlocking Business Potential in FMCG with Predictive Analytics: A Machine Learning Approach

Sharma, Utkarsh (2024) Unlocking Business Potential in FMCG with Predictive Analytics: A Machine Learning Approach. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (992kB) | Preview
Preview	PDF (Configuration Manual) Download (905kB) | Preview

Abstract

The Fast-Moving Consumer Goods (FMCG) segment is one of the most important segments in the world economy comprising goods that can be sold quickly at rather low margin – foods & beverages, non food personal healthcare products, home care etc. Traditional models of the FMCG segment have burdened organizations with the problems of low sales forecast, customer actions examination, and stock manage. Many of these models are based upon historical sales data and last tendencies plus simple forecasting methods which do not reflect current market or consumer demands, seasonality, and variability of many aspects of supply chain. These challenges are addressed in this study by using advanced machine learning (ML) models and algorithms for decision-making process in the FMCG operations. Using the tools of predictive sales, customer behavior analysis using K-Means clustering and optimisation of an inventory through reinforcement learning, the paper gives a more comprehensive approach. In this paper, the authors highlight an integrated system for supply chain management and sales forecasting with the assistance of ML methods. The study focuses on three primary domains: forecasting of sales, understanding of the customer behaviour and the management of supply chain efficiently. In the first domain, five models of ML Linear Regression, Random Forest, Decision Tree, SVR, KNN was used to forecast the retail sales to which the Random Forest was exclusively identified as the most suitable model within the three performance indicators of RMSE as well as R². The second domain used K-Means clustering to assess customers’ behavior; this broke down the customers into three groups based on their retail, transfer and warehouse sales. These segments offered valuable information when it comes to the management of stock and sales outlets. The third domain was concerned with stock-up or stock-down decisions based on seasonal sales forecasts of the organisation’s products. Moreover, for the formation of an inventory control system, Reinforcement Learning was used to find the optimal between stock surplus and stockout, holding cost, and profit.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Agarwal, Bharat UNSPECIFIED
Uncontrolled Keywords:	Fast-Moving Consumer Goods (FMCG); Sales Prediction; Machine Learning (ML) Models; Random Forest; K-Means Clustering; Customer Behavior Analysis
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HF Commerce > Marketing > Consumer Behaviour H Social Sciences > Economics Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	05 Sep 2025 08:55
Last Modified:	05 Sep 2025 08:55
URI:	https://norma.ncirl.ie/id/eprint/8808

Actions (login required)

View Item