PERSONALIZED AD TARGETING SYSTEM USING MACHINE LEARNING FOR OPTIMIZING AD DELIVERY

Abstract

In the modern digital marketing landscape, personalization has become a critical factor for successful advertising, with over 80% of users favoring personalized ads and such campaigns yielding up to 20% higher conversion rates than generic approaches. Traditional ad-targeting methods relied heavily on manual analysis of customer records, historical sales, and viewership data. These techniques were inefficient, prone to human error, and lacked the ability to adapt in real time, often resulting in suboptimal campaign performance. Early machine learning models, such as the Gradient Boosting Classifier, improved prediction accuracy compared to manual methods but suffered from limitations including overfitting, slow training on large-scale data, and poor adaptability to rapidly evolving user preferences, restricting their effectiveness in real-time personalization. To address these challenges, we propose a hybrid machine learning framework that integrates a Feedforward Neural Network (FNN) with an Extra Trees Classifier. The FNN effectively captures complex, non-linear patterns in user behavior, while the Extra Trees Classifier provides efficient feature selection and fast, robust predictions. By leveraging user data such as browsing history, click-through rates, demographic attributes, and prior ad interactions, the system generates highly personalized ad recommendations. Experimental results demonstrate that the proposed model achieves a classification accuracy of 91%, representing a 28% improvement over Gradient Boosting, along with a notable reduction in training time. Overall, this adaptive and scalable ML pipeline enhances ad targeting precision, improves user engagement, and maximizes return on ad spend in dynamic digital environments