The protection problems in microgrid effect the reliability of the power system caused due to high distributed generator penetrations. Therefore, fault protection in microgrid is extremely important and needs to be resolved to enhance the robustness of the power system. This manuscript proposes a combined signal processing and data mining-based approach for microgrid fault protection. In this study, first the multiresolution decomposition of wavelet transform is employed to preprocess the voltage and current signals to compute the total harmonic distortion of the voltage and current. Then, the statistical indices of standard deviation, mean, and median of the total harmonic distortion and the negative sequence components of active and reactive power are used to collect the input data. After that, all the available data is provided to the random forest-based classifier to evaluate the efficiency of the proposed scheme in terms of the detection, identification, and classification of faults. This study used different aspects for data collection by simulating various fault and no-fault cases for both looped and radial configurations under grid-connected and islanded modes of operation. The simulations were performed on a standard medium voltage microgrid using MATLAB/SIMULINK, whereas the analysis for testing and training of the random forest were conducted in Python. It is recognized that the proposed method performed better than support vector machines and decision tree that are reported in the literature. The results further demonstrate that the proposed method can also detect simultaneous faults, and it is also effective against measurement noise.