Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

Click Through Rate Prediction Leveraging Machine Learning Techniques for Mobile Digital Advertisement

Rojas Guillen, Juliana Margaret LU (2024) DABN01 20241
Department of Economics
Department of Statistics
Abstract
Predicting click-through rates (CTR) is essential for optimizing the effectiveness of mobile advertising campaigns, where accurate prediction of user interactions can significantly enhance revenue generation and ad targeting strategies. This thesis investigates the efficacy of different predictive models, using a dataset composed of impressions and interactions with mobile ads. The models examined include Logistic Regression, Random Forest, XG-Boost, CatBoost, and Feed Forward Neural Networks. Additionally, a K-Means Clustering approach was employed to segment the data into clusters prior to modeling. The findings reveal that ensemble methods, particularly CatBoost, outperformed the other tested models, delivering the lowest log-loss... (More)
Predicting click-through rates (CTR) is essential for optimizing the effectiveness of mobile advertising campaigns, where accurate prediction of user interactions can significantly enhance revenue generation and ad targeting strategies. This thesis investigates the efficacy of different predictive models, using a dataset composed of impressions and interactions with mobile ads. The models examined include Logistic Regression, Random Forest, XG-Boost, CatBoost, and Feed Forward Neural Networks. Additionally, a K-Means Clustering approach was employed to segment the data into clusters prior to modeling. The findings reveal that ensemble methods, particularly CatBoost, outperformed the other tested models, delivering the lowest log-loss (0.5836) and the highest F1-score (0.7093). This superior performance highlights the robustness of gradient-boosting machines in handling mobile ad data, which is often categorical and highly dimensional. Finally, features such as site_id, app_id, and device_model were identified as the most influential in the prediction of CTR using the best-performing model. (Less)
Please use this url to cite or link to this publication:
author
Rojas Guillen, Juliana Margaret LU
supervisor
organization
course
DABN01 20241
year
type
H1 - Master's Degree (One Year)
subject
keywords
Click-Through Rate, Machine Learning, Mobile Ads, CatBoost
language
English
id
9155009
date added to LUP
2024-09-24 08:36:17
date last changed
2024-09-24 08:36:17
@misc{9155009,
  abstract     = {{Predicting click-through rates (CTR) is essential for optimizing the effectiveness of mobile advertising campaigns, where accurate prediction of user interactions can significantly enhance revenue generation and ad targeting strategies. This thesis investigates the efficacy of different predictive models, using a dataset composed of impressions and interactions with mobile ads. The models examined include Logistic Regression, Random Forest, XG-Boost, CatBoost, and Feed Forward Neural Networks. Additionally, a K-Means Clustering approach was employed to segment the data into clusters prior to modeling. The findings reveal that ensemble methods, particularly CatBoost, outperformed the other tested models, delivering the lowest log-loss (0.5836) and the highest F1-score (0.7093). This superior performance highlights the robustness of gradient-boosting machines in handling mobile ad data, which is often categorical and highly dimensional. Finally, features such as site_id, app_id, and device_model were identified as the most influential in the prediction of CTR using the best-performing model.}},
  author       = {{Rojas Guillen, Juliana Margaret}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{Click Through Rate Prediction Leveraging Machine Learning Techniques for Mobile Digital Advertisement}},
  year         = {{2024}},
}