The unlawful usage of electricity without paying for it is termed electricity theft. The theft of electricity can occur in three main ways—sudden high consumption, energy theft, and defective meters. Electricity theft leads to financial crises causing economic disruption in developing countries. Machine learning techniques are introduced to understand the energy consumption captured regularly by the meters, identifying any abnormalities in the usual consumptions, detecting any malicious activities of the meters, and also spotting defect meters. This research work understands the data and identifies electricity theft as a classification problem. The comparative analysis of the five classifiers is carried out, their performance is evaluated in terms of TPR against FPR curves. In order to standardize the unbalanced data, interpolation techniques are used. Consequently, KNN, RF, and XGB classifiers perform better than other classifier models. XGB obtained 93% accuracy and 98% AUC score, which was better than KNN and RF.