Electricity theft poses a significant challenge to power utilities, resulting in substantial financial losses and compromising the reliability of the power system. Effective Electricity Theft Detection (ETD) is crucial for minimizing financial losses and ensuring the economic operation of the power system. This paper proposes a predictive ETD framework based on consumer load profiles. Statistical features, such as the mean and standard deviation of monthly energy consumption, are used to distinguish between normal and abnormal usage patterns. Several machine learning algorithms are applied to classify consumption behavior, and the model's performance is evaluated using metrics such as accuracy, precision, recall, and the Receiver Operating Characteristics (ROC). The results demonstrate that the proposed feature-based approach enhances detection accuracy, validating the framework's effectiveness in identifying electricity theft.