Reservoir porosity determines how much fluid formation can store, making its accurate estimation a key element of petrophysical evaluation, reservoir characterization and geological studies. Core analysis provides reliable readings, but covers limited intervals and time-consuming, expensive. Log-based equations derived from well-logging interpretation introduce uncertainties that affect prediction reliability. Four ensemble machine learning models were used here to enhance porosity prediction, Decision Trees (DT), Random Forests (RF), Gradient Boosting (GB) and Extreme Gradient Boosting (XGB). Input dataset used 734 data point from seven wells in a Libyan field while a separate well with 202 data point kept as a blind well to evaluate model generalization. Bulk density (RHOB), gamma ray (GR), compressional travel time (DT), and neutron porosity (CNL) chosen as input features because each shows established relationship with porosity. Statistical metrics including correlation coefficients (R2) and root-mean squared error (RMSE) were used to assess the model. The results indicate that ensemble models provide a robust and efficient alternative to conventional porosity estimation methods, offering improved predictive accuracy and reliability. This study highlights the potential of machine learning in reservoir characterization, contributing to more data-driven decision-making in petroleum engineering applications. This scientific manuscript was presented at the sessions of the International Renewable Energy, Gas, Oil and Climate Change Conference "iREGO" in the period of April 25-27, 2026. Tripoli - Libya................. Keywords:............ Porosity Prediction, Ensemble models, Machine Learning, Well Logging Data.