Introduction

A sophisticated approach to identify fraud electricity consumption from customers were developed. The consumption data from customers was used to train the LSTM algorithm. But this input data was sliced by season, by day, by month and even by hour in order identify several electricity use patterns. For this task different data sources were integrated with python data processing code, Tensorflow data analytics framework and Dash framework for dynamic visual representation.

The final result is an accurate and self explanatory graphical fraud detector tool with dynamic graphs.