This paper proposes a score based computational technique for the detection of non-technical losses in electricity distribution networks. The methodology is comprised of three steps. In the first one, a score is assigned to each meter number considering the area that customers live. In second step, a Cmeans-based fuzzy clustering is applied to find consumers with similar consumption profiles. Then, a fuzzy classification is performed with fuzzy membership matrices. Afterwards, the Euclidean distances between membership matrices are calculated and normalized, yielding an index score. In third one, expected consumption values of each customer are calculated with installed power values and compared with real usage values. The differences are used as another score. Using all scores, a final score has been formed for each consumer, to be used to detect potential fraudsters. The approach was tested and validated on a real dataset, showing good performance in tasks of abnormal usage detection.