End-of-life product recycling is crucial for achieving sustainability in circular supply chains and improving resource utilization. Forecasting the quantity of recycled end-of-life products is essential for planning and managing reverse supply chain operations. Decision-makers and practitioners can benefit from this information when designing reverse logistics networks, managing tactical disposal, planning capacity, and operational production. To address the challenge of small sample data with multiple factors influencing the recycling number, and to deal with the randomness and nonlinearity of the recycling quantity, a hybrid predictive model has been developed in this research. The model is based on k-nearest neighbor mega-trend diffusion (KNNMTD), particle swarm optimization (PSO), and support vector regression (SVR) using the data from the field of end-of-life vehicles as a case study. Unlike existing literature, this research incorporates the data augmentation method to build an SVR-based model for end-of-life product recycling. The study shows that developing the predictive model using artificial virtual samples supported by the KNNMTD method is feasible, the PSO algorithm effectively brings strong approximation ability to the SVR-based model, and the KNNMTD-PSO-SVR model perform well in predicting the recycled end-of-life products quantity. These research findings could be considered a fundamental component of the smart system for circular supply chains, which will enable the smart platform to achieve supply chain sustainability through resource allocation and regional industry deployment.