| Abstrakt: |
Brazilian Energy Distributors need to pay special attention to electric energy loss due to non-technical irregularities. Fraud and robbery represents a loss of revenue of billions of dollars every year. The energy utilities first approach to this problem was the creation of campaigns for prevention, raising awareness of the population and encouraging the denunciation of fraudsters. Unfortunately, this was not enough and proactive actions were needed. The current effective strategy is conducting technical inspections at consumers to whom there is some kind of evidence, based on customer data, consumption patterns, reader's notes, etc. This paper the application of neural network models, nnet and avNNet, on selecting a set of consuming units for which there is a high probability of frauds and irregularities. It also includes data enrichment to enhance the results of such neural networks. First, it presents a brief introduction about the problem, scope of this article and related work. The data, to which the companies usually have access, is analyzed and the classification model construction is described. Then, the resulting models are compared considering the precision, accuracy and duration of training. Finally, there is a reasoning concerning such results and remarks about how energy distributors may use this information. [ABSTRACT FROM AUTHOR] |
