Photovoltaic (PV) panels can encounter various issues due to an aggressive working environment and malfunctions caused by environmental or human factors. These issues, known as faults, can significantly reduce PV system performance and increase the risk of fire. To address this, numerous diagnostic methods have been developed, with fault classification being a principal part of diagnosing faults in PV panels. This article proposes a fault classification approach using two machine learning algorithms, with I(V) characteristics as input parameters. These algorithms are designed to identify three specific faults in the series-parallel configuration of solar panels: partial shading, bypass diode short circuits, and ground fault. These faults are commonly found in photovoltaic fields. Data preprocessing techniques are employed to enhance the quality of the fault dataset. A comparison is made between two algorithms, considering factors such as accuracy, Standard deviation of scores, root mean square error (RMSE). Experimental results show that the diagnostic algorithm based on random forest (RF) exhibits higher precision and generalization capability compared to k-Nearest Neighbors (kNN) for classifying faults in photovoltaic modules within a solar field.