The oxidative potential (OP) of particulate matter (PM) is widely recognized as an important measure for evaluating the risk of oxidative stress imbalances caused by air pollution. Despite its significance, the primary sources of OP in different regions and the interactions among these sources remain unclear. This study investigates OP sources and their interactions using PM_2.5 samples from Suzu, Japan, collected during 2015–2016. Chemical composition analyses were conducted and positive matrix factorization (PMF) was applied to identify the sources of PM_2.5 in Suzu. The dithiothreitol (DTT) assay was used to quantify the OP of PM_2.5. To capture the nonlinear behavior of atmospheric processes, a multilayer perceptron (MLP), a type of artificial neural network, was utilized to model OP contributions from the PMF-identified sources. The constructed models were then applied to simulate the interactions between the pollution sources in different seasons by modifying their input intensities. The results indicate that while the contributions of PM_2.5 sources to OP are generally within the range of additive effects, there is a possibility of significant interactions between certain PM_2.5 sources, whether synergistic or antagonistic, which deviate from the additive assumptions of traditional linear models. Furthermore, some interactions intensify as PM_2.5 concentrations increase. These findings underscore the necessity of accounting for source interactions and call for detailed monitoring of various PM_2.5 sources to accurately assess OP risks and develop effective air pollution management strategies.