Fraud detection in energy distribution using support vector machines

Several economics sectors are exposed to fraud made by their own customers. At the Electrician Distribution segment is not different. Several techniques in statistical fields were developed to detect illegal activities, relied in observation classification. The empiric events modeling always become a challenge to get solution in engineering projects. The solution is achieved using a induction process to built a system able to bring up the answer of a previously observed event. The Linear Discriminate Analysis is the quantitative method most used. Recently, an alternative approach arises: The Support Vector Machine (SVM). This paper objectives is train and test a model built using SVM to point out the customers that are performing frauds given the customers company (an Electrician Distribution) data base, doing a efficiency and quality confrontation vis–à–vis the Linear Discriminate Analysis.