Disposal of plastic waste causes serious environmental problems, including landfills and water bodies degradation, greenhouse gas emission, soil contamination and so on. This study investigates the use of recycled plastic aggregate (RPA), as a partial replacement for conventional coarse aggregates by weight in lightweight concrete production. Concrete mixtures with different concentrations of RPA at (0, 15, 30, and 45%) were prepared and cured for 7, 14, 28, and 56 days. Including RPA into concrete reduced both density and compressive strength as the replacement level increased. Density decreased from 2,347 kg/m³ at 0% RPA to 1,895 kg/m³ at 45% replacement. Similarly, 28 days compressive strength decreased from 30.43 N/mm² (control) to 19.50 N/mm² at 45% replacement, reflecting the lower specific gravity and weaker bonding of RPA compared to traditional coarse aggregate. Additionally, the test results showed that RPA concrete has a low water absorption rate at 15% replacement, with 2.50% for water absorption and a 0.0235 mm/s^1/2 sorptivity value compared to control samples with 2.66% for water absorption and 0.024 mm/s^1/2 sorptivity value. However, concrete samples with up to 30% RPA replacement met the minimum requirements for structural lightweight concrete. This study also used machine learning models, including artificial neural networks (ANN), k-nearest neighbor (k-NN), and random forest (RF), to predict the durability properties of RPA concrete. Among these models, the k-NN model showed the best prediction accuracy with an R² value of 1.00, a mean absolute error (MAE) and a mean square error (MSE) of 0.001 for both the train and test data. These findings show that the use of treated RPA in concrete not only offers a sustainable alternative to natural aggregates but also improves the durability of the resulting structures.