Non-technical energy losses represent a significant challenge for electric utility companies, as they are attributable to fraud such as illegal connections or meter tampering. This problem generates enormous costs and is extremely difficult to detect and prevent. To address the problem, the equipment used to measure in the process layer of Industry 4.0 expands and updates, giving access to a large amount of data that are obtained frequently and need to be processed and analyzed. In this paper we propose a data analytics methodology for the detection of non-technical losses with machine learning and big data techniques.