Predicting household electricity consumption using data mining techniques -Deep neural networks LSTM-

Abstract

The study addressed the topic of predicting electricity consumption using deep neural networks, which are one of the most effective methods in the field of data mining, as they process huge data with a time sequence. The study focused on the electricity consumption data of the residents of Ghardaïa State and identified the most important electricity tariffs that allow for rationalizing consumption and reducing the cost of the electricity bill.

The study concluded that deep neural networks play an effective role in providing a future picture of electricity consumption from 2023Q4 to 2025Q4, with an explanatory power coefficient  of 75% and a mean square error (MSE) of 25%. This means that the difference between actual and predicted values ​​is small, resulting in higher accuracy.