Detecting and mitigating Non-Technical Losses (NTL’s) , poses a significant challenge to electric utility companies, and NTL’s increase the cost of operating a power system. NTL’s are calculated by the overall difference between the produced energy at the generation levels and the collected household electricity consumption bills. This difference is considered a waste of the natural energy resources used in generating electricity. Due to the complexity of power systems and the electricity generation process, where there are many coupled factors that may result in anomalies in consumption bills, the detection of NTLs is not a trivial task. Such factors include the naturally occurring losses in power systems, technical failures, and consumer behavior changes due to tariffs and electricity prices changes. However, by using the right techniques of artificial intelligence and machine learning, it is possible to detect NTLs with high accuracy and automate accounts inspection processes that reduce system operational costs while improving the use of inspection resources. By integrating analysis techniques that model normal and fraudulent electricity consumption patterns, in addition to engineering many relevant features that aid in the classification task, we produced highly accurate classification models that detect NTLs with high confidence. The models predicted the likelihood of an NTL for each electrical meter and ranked the suspect accounts based on the potential of retrieved losses. The model also considered the optimum allocation of inspection resources and the spatial and socioeconomic correlations of NTLs.