This work proposes a Machine Learning-based Prioritized-Recommendation System for the retail industry. Traditional recommendation systems often face challenges in ranking items effectively, which can lead to lower engagement and sales. This research addresses this gap by integrating Association Rule Mining (ARM) with a Random Forest classifier to predict the likelihood of customer purchases based on transaction data. The system first extracts item relationships using ARM and creates feature vectors from these rules and customer data. A Random Forest model is then used to predict purchase probabilities, and prioritization ranks recommendations accordingly. The proposed system archives model accuracy of 94.96%, Precision 0.85, Recall 0.893, F1-Score 0.88, and AUC-ROC 0.93, demonstrating superior performance in recommending the most relevant products. Furthermore, the use of Incremental ARM ensures the model can dynamically adapt the changes in customer behavior, providing an efficient and scalable solution for personalized retail marketing.