Partial shading affects the performance of photovoltaic (PV) systems, as it reduces the performance and reliability of the system. A smarter PV panel that can reduce the impact of partial shading conditions through machine learning (ML) and transistor-embedded PV panels is desirable for a wider acceptance of residential PV panels. To achieve it, this paper presents an ML-based approach to classify the shading conditions of PV panels to enable real-time detection and reconfigurability of transistor-embedded PV panels. A dataset consisting of over 24 million datapoints was generated using a MATLAB Simulink model of a 60-cell PV panel operating under varying environmental conditions, including irradiance, temperature, and shading levels. Features such as voltage, current, power, and engineered ratios are used to train and evaluate five ML classifiers: Logistic Regression, k-Nearest Neighbors (KNN), Random Forest (RF), XGBoost, and LightGBM. Among these, ensemble-based model Random Forest demonstrated superior performance, achieving a classification accuracy of 97.5% and macro F1-score of 96.4%. A detailed analysis of precision, recall, F1 score, and importance of characteristics reveals that the design features and configuration parameters significantly influence classification accuracy. The proposed methodology holds promise for enabling adaptive control in smart PV panels for residential applications.