<![CDATA[Predicting online shoppers' purchase intention is a key issue in e-commerce because it directly affects conversion and marketing effectiveness. The main focus of this article is a Random Forest purchase-intention model accompanied by an end-to-end MLOps implementation to ensure production readiness. The dataset used is Online Shoppers Intention with 12,330 samples and 18 features representing administrative, informational, and product-related characteristics, along with behavioral metrics. Preprocessing includes missing-value imputation, numerical feature standardization, categorical feature encoding, and outlier removal using the z-score method. The model is optimized with GridSearchCV and 3-fold cross-validation. Test results show 91.38% accuracy with 73.60% precision, 56.64% recall, and 64.02% F1-score for the positive class. MLOps implementation uses MLflow for experiment tracking, Prometheus-Grafana for monitoring, and a GitHub Actions-based CI/CD pipeline for deployment automation. Overall, the Random Forest model delivers strong predictive performance on e-commerce data and is supported by an MLOps pipeline that improves reproducibility, deployment, and production monitoring]]>
Copyrights © 2026
