Dynamic Advertisement Pricing and Bidding Optimization: An Integrated Machine Learning and Auction Framework

Online advertising platforms rely on efficient pricing and bidding strategies to maximize returns for advertisers while maintaining fairness and competitiveness in auctions. This paper presents an integrated machine learning and auction-based framework that combines Click-Through Rate (CTR) prediction, convex optimization, fairness evaluation, and multi-agent auction simulation. Synthetic datasets mimicking real-world ad environments were generated, and multiple models, including Logistic Regression, Random Forest, and XGBoost, were evaluated. XGBoost achieved the highest ROC-AUC (0.8731) and lowest log loss (0.2187), improving F1-score by 12% over the baseline after applying SMOTE. Predicted CTRs were used in a convex optimization model, solved with CVXPY, to allocate budgets optimally, increasing ROI by up to 23.5% compared to uniform bidding. A multi-agent second-price auction simulation demonstrated that balanced bidding strategies improved clicks-per-dollar efficiency by 22% over aggressive bidding. Fairness analysis across gender groups revealed minimal disparity,with a prediction accuracy gap of only 1.3%. Comparative evaluation against LightGBM, CatBoost, and heuristic baselines confirmed the superiority of the proposed method in both prediction and ROI. The proposed framework is computationally efficient, scalable, and applicable to diverse online advertising scenarios.