Tools
Tools
Atharva, Atharva Sachin (2026) A Multi-Model Machine Learning Framework for Analysing Conversion Behaviour in Digital Advertising. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Abstract
The increasing complexity of user behaviour and the rapid increase in digital advertising platforms have led to the large-scale clickstream data, posing large challenges to meaningful optimisation of a campaign. Conventional descriptive analytics, which are based on the measures of impressions and clicks, are not always sufficient to capture intricate behavioural trends or reliably predict the conversion rates. This has consequently led to a situation where machine learning is now a very critical method of deriving actionable information regarding advertising data. In this research, a machine-driven prediction of conversion and performance estimation in the sponsored search advertisement based on the real-world, click-level monitoring data are postulated. The purposes are to determine the main factors that determine conversion behaviour, properly categorise conversion outcomes, subdivide behavioural patterns and predict short term conversion trends. In order to achieve these goals, a lot of data preprocessing was done, such as missing value processing, numerical feature scaling, high-cardinality categoricals encoding, and feature engineering. A number of supervised learning models were developed and tested, which are Logistic Regression, Random Forest, Gradient Boosting, and LightGBM. Logistic Regression was used as a control group and it had a 93.5% accuracy. This baseline was much less effective as compared to ensemble-based models. Random Forest model performed almost perfectly on 100% precision, recall, and F1-score and almost on the ROC-AUC value of approximately 1.00 whilst Gradient Boosting and LightGBM scored the F1-score at approximately 0.96. Furthermore, K-Means clustering also found behavioural segments. Altogether, the findings prove that ensemble machine learning methods are effective to model multifaceted advertising data and aid campaign optimisation by using data.
| Item Type: | Thesis (Masters) |
|---|---|
| Supervisors: | Name Email Haque, Rejwanul UNSPECIFIED |
| Uncontrolled Keywords: | Digital Advertising Analytics; Machine Learning; Campaign Performance Prediction; Time-Series Forecasting; Prophet Model; Google Ads; Meta Ads |
| 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 > Advertising H Social Sciences > HF Commerce > Marketing > Consumer Behaviour 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: | 09 Nov 2026 19:10 |
| Last Modified: | 09 Nov 2026 19:10 |
| URI: | https://norma.ncirl.ie/id/eprint/9952 |
Actions (login required)
 |
View Item |