A study was conducted to predict future household gas consumption for 100,000 consumers across different time intervals utilizing machine learning algorithms. Furthermore, neural networks were employed to identify key consumption trends and detect any abnormal patterns or fraudulent activity. The algorithms utilized in the study encompassed a range of techniques including ARIMA, Prophet, LSTM, EDA, Linear Regression, Lasso, Ridge, K-Neighbors Regressor, Decision Tree, Random Forest Regressor, XGBRegressor, CatBoost Regressor, and AdaBoost Regressor.Upon the conclusion of the project, it was discovered that among a sample of 100,000 consumers, around 5% were identified as heavy consumers, and approximately 1% were characterized as delinquent consumers.

In the initial stages, it is important to note that this project involves compiling data from various stations across different provinces and cities, specifically focusing on information relevant to the main branches for transfer to towns, offices, or factories. The primary objective of this project is to utilize machine learning techniques to forecast energy demand. This will involve analyzing historical data pertaining to energy production sources, weather patterns, and other relevant variables in order to develop an accurate forecasting model. Key tasks within the project scope include data preprocessing, exploratory data analysis, feature engineering, model training, and evaluation.The dataset consists of details on electricity generation from diverse sources, actual load information, and weather data encompassing temperature, pressure, humidity, wind speed, and cloud cover. Ultimately, the project aims to pinpoint stations exhibiting high levels of energy consumption.