Gearboxes are commonly employed in rotating machinery, particularly in wind turbines, to transmit power and torque. Their possibility of failure may be increased by the extended hours they work and the various working situations they encounter. A sudden gearbox failure might result in severe downtime and increased maintenance expenses. Because of this, a rising percentage of wind turbines are outfitted with vibration measurement devices to facilitate condition monitoring, which seeks to identify potential problems early on and enhance maintenance. The integration of machine learning algorithms enables the prevention of degradation of any element present in a wind turbine, as well as the identification and diagnosis of unanticipated failures. This paper presents a study on National Renewable Energy Laboratory (NREL) wind turbine gearbox fault classification accuracy by employing various Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) machine learning classifiers, with an emphasis on the effect of different Spectral Estimation Methods. Using Welch’s Method, Autoregressive Model, and State-Space Model, the power spectrum was created, and for each spectrum, the frequency domain features were extracted simultaneously. Finally, by analyzing the performance of the different classifier, the effective spectral estimation method among Welch’s Method, Autoregressive Model, and State-Space Model was determined.