Abstract
Theft of electricity is a problem in many developing countries. But AMI is paving the way for data-centric ar-
chitecture to help in theft detection. However, a smart grid or even AMR is a long shot for many developing
countries due to the costs involved in its large-scale deployment. This paper presents a technique to detect
outliers among electricity users that further investigates electricity theft using data analytics on monthly usage
data available to every utility company. Using this technique, we have reduced the search space for theft
identification to as low as 3.4% of the total customer base.