MustafaAssem2024/Predicting-Power-Consumption-in-Tetouan-City-Morocco

This project aims to predict power consumption in Tetouan City, Morocco, using historical data and advanced machine learning techniques. Predicting Power Consumption in Tetouan City, Morocco Overview This project focuses on predicting power consumption in Tetouan City, Morocco, through machine learning models. The work is divided into two main parts: Reproducing Original Results: The first part replicates the findings from a study on predicting power consumption in Tetouan City using similar features, models, and parameters. This helps validate the original results. Developing a New Solution: The second part presents an alternative solution that aims to improve or match the original study's performance using new feature selection methods, model optimizations, and innovative machine learning algorithms. Project Structure Part 1: Reproducing Original Results Followed the methods detailed in the original study to replicate the results. sosurce: ieeexplore.ieee.org Used various models, including Linear Regression, Decision Tree, Random Forest, Support Vector Regression (SVR), and Feedforward Neural Network (FFNN). Results were compared against the original study to assess accuracy and consistency. Part 2: Developing a New Solution Designed an ensemble model approach to improve predictive accuracy. Implemented techniques such as stacking predictions from multiple models (Gradient Boosting Regressor, Random Forest, FFNN) and using Ridge Regression as the final model. Demonstrated the ensemble model’s performance, particularly in Zones 2 and 3, where it significantly outperformed the original study's models. Dataset Data: The dataset includes hourly power consumption data for three different zones in Tetouan City, along with environmental variables such as temperature, humidity, and wind speed. Preprocessing: The dataset was cleaned, and features were engineered, including datetime extraction to create new time-related features. Models Used Part 1: Reproducing Original Results Linear Regression Decision Tree Random Forest Support Vector Regression (SVR) Feedforward Neur …