Comparing Statistical and Machine Learning Models for Forecasting Milk Consumption in Ireland

Satpute, Utkarsh Arun (2025) Comparing Statistical and Machine Learning Models for Forecasting Milk Consumption in Ireland. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (871kB) | Preview
Preview	PDF (Configuration Manual) Download (441kB) | Preview

Abstract

Accurate forecasting of dairy consumption is essential for effective supply chain management and policy planning for Ireland. From this research we will aims to explore and evaluate the performance of various time series forecasting models in predicting milk consumption trends across multiple categories, which include whole milk, skimmed milk, and semi-skimmed milk Irish dairy data. The goal is to identify the most accurate forecasting model for each category by comparing traditional statistical approaches with modern machine learning and deep learning techniques.

To acquire the result, we have implemented a data preprocessing technique in which it includes handling of missing values, stationarity, seasonal decomposition and testing of normalisation, and to evaluate the models we have used three type of evaluation metrics which are the (RMSE, MSE, R²). We also did an comparative analysis of raw versus normalized data revealed notable differences in model behaviour. The results highlight varying strengths of each model depending on the milk category and data transformation was applied.

Among the tested models, SARIMA performed best for the whole milk and for the import and export dataset. While XGBoost outperformed others on skimmed milk. And one deep learning model that is LSTM achieved the strongest performance for the Semi – Skimmed milk. But other deep learning model Prophet underperformed on the irregular series. We also applied normalization which improved ML/DL models, but it reduced statistical accuracy of the models.

This study will contribute to a benchmark for forecasting techniques and offers practical insights for dairy producers and policymakers. The research concludes with a discussion on model selection, limitations and recommendations for future work, including deeper feature engineering and real-time predictive systems

Item Type:	Thesis (Masters)
Supervisors:	Name Email Kumar Menghwar, Teerath UNSPECIFIED
Subjects:	H Social Sciences > HD Industries. Land use. Labor > Specific Industries > Agriculture Industry H Social Sciences > HD Industries. Land use. Labor > Specific Industries > Food Industry > Beverage industry D History General and Old World > DA Great Britain > Ireland Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning H Social Sciences > HD Industries. Land use. Labor > Business Logistics > Supply Chain Management
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	26 Aug 2026 11:40
Last Modified:	26 Aug 2026 11:40
URI:	https://norma.ncirl.ie/id/eprint/9665

Actions (login required)

View Item