<![CDATA[Indonesia's oil and gas industry faces ongoing challenges in maintaining production due to the maturity and decline of many fields. Success in drilling new wells is vital for sustaining output, but performance forecasting is hampered by reservoir variability, geological complexity, and operational differences. These challenges underscore the need for adaptive, data-driven predictive methods in technical planning. This study develops a machine-learning-based model for predicting Estimated Ultimate Recovery (EUR) using a KNIME workflow. K-Means clustering groups wells by reservoir characteristics, and three regression algorithms (Simple Regression, Gradient Boosting, and Random Forest) are compared for predicting EUR. The combined workflow evaluates how reservoir segmentation improves EUR prediction accuracy. The methodology consists of four main stages: (1) collecting and preprocessing historical well data from long-producing oil fields located in the Pekanbaru region, (2) applying K-Means clustering using reservoir features such as porosity, permeability, Net Pay, Water Saturation, and well coordinates, (3) constructing EUR regression models using Simple Regression, Gradient Boosting, and Random Forest with an 80% training and 20% testing scheme, and (4) validating model performance using evaluation metrics such as R² and RMSE. All processes were performed using the KNIME platform to ensure a standardized, transparent, and easily replicable workflow. This study is expected to produce an EUR prediction model that is accurate and stable, while also identifying the most suitable regression algorithm for mature Indonesian reservoirs. Furthermore, the integrated KNIME workflow can serve as a foundation for a decision support system to assist in drilling planning, investment optimization, and the reduction of production uncertainties. In addition, the results of the clustering and modeling can be used to identify the best prospective reservoir zones, thereby supporting the selection of new well drilling locations in areas with the highest production potential.]]>
