This article introduces the design of a substrate integrated waveguide (SIW) resonator-based microwave sensor, integrated with a multilayer perceptron (MLP) neural network, for the accurate classification and quantification of water contaminants in fuel oil.The system utilizes key features derived from the microwave sensor output, including the magnitude of the transmission coefficient (S21) and the resonance frequency (Fr), to train the MLP network.In the development of the sensor, the prototype was able to effectively monitor water contamination in fuel oil, spanning concentrations from 0 % to 100 %, by filling a cylindrical tube positioned at the sensor′s center.Operating within the frequency range of 1.50-1.55 GHz, the sensor demonstrated robust performance.The study further explores the correlation between water contamination levels and sensor response using both linear and nonlinear regression analyses, complemented by the application of an MLP neural network for contaminant classification.The proposed sensor achieves non-contact measurement of water contaminants with sensitivities of 0.12 dB/% and 0.88 MHz/%.Given its high sensitivity, real-time monitoring capabilities, user-friendly design, cost-effectiveness, and non-invasive approach, the sensor holds significant promise for applications in the petroleum industry.