Accurately predicting reservoir porosity and permeability is of great significance for evaluating the storage capacity of tight sandstone reservoirs. Traditional methods for petrophysical evaluation often yield significant errors. To fully utilize existing well logging data and improve prediction accuracy, a literature review revealed that conventional methods have limitations and perform poorly in predicting the petrophysical parameters of tight sandstone reservoirs. Therefore, this paper proposes a model that integrates the Particle Swarm Optimization (PSO) algorithm to globally iteratively search for the optimal penalty factor (c) and kernel function parameter (g) for a Support Vector Machine (SVM) prediction model. This PSO-SVM model is constructed for predicting the petrophysical parameters of tight sandstone reservoirs. In practical application, Pearson correlation coefficients were calculated to select well logging curves highly sensitive to porosity and permeability. Five curves—Spontaneous Potential (SP), Gamma Ray (GR), Caliper (CAL), Acoustic (AC), and Compensated Neutron Log (CNL)—were chosen as input features to predict porosity and permeability, respectively. Research demonstrates that the PSO-SVM model achieved prediction accuracies of 96.7% for porosity and 93.6% for permeability, outperforming other similar algorithms. It proves to be a reliable and advantageous method for predicting petrophysical parameters in the tight sandstone reservoirs of the Shanxi Formation in Block Q, providing robust technical support for oil and gas exploration and development.