⚡ Electric Power Consumption Prediction using Random Forest

This project predicts household electric power consumption using the UCI Individual Household Electric Power Consumption Dataset.
The goal is to accurately forecast Global Active Power based on features such as voltage, reactive power, current, and sub-meter readings.

---

📘 Project Overview

This notebook demonstrates the full pipeline of a Machine Learning regression project:

1. Data Loading – Imported directly from the UCI Machine Learning Repository
2. Data Cleaning – Handled missing values, replaced '?' with NaN, interpolated time-series data
3. Feature Engineering – Combined date and time columns, removed outliers, and converted all data to numeric types
4. Model Training – Used Random Forest Regressor for prediction
5. Model Evaluation – Evaluated using metrics like R², MAE, MSE, and MAPE
6. Visualization – Compared actual vs predicted power consumption using scatter plots and line charts

---

🧠 Key Features

• Cleaned and preprocessed 1.9 million energy data points
• Used Random Forest for robust and high-accuracy regression
• Achieved R² = 0.998 and Average Accuracy ≈ 96.7%
• Generated performance visualizations:
  • Actual vs Predicted Scatter Plot
  • Residual Distribution
  • Feature Importance Plot

---

📊 Dataset Information

Dataset Name: Individual Household Electric Power Consumption
Source: UCI Machine Learning Repository
Rows: ~2 million
Columns:

• Global_active_power — Total active power consumed (Target)
• Global_reactive_power — Reactive power
• Voltage — Average voltage
• Global_intensity — Average current
• Sub_metering_1, Sub_metering_2, Sub_metering_3 — Energy consumption in different household areas

---

🧰 Tech Stack

• Python 3
• Pandas, NumPy — Data manipulation
• Scikit-learn — Model training and evaluation
• Matplotlib, Seaborn — Visualization
• ucimlrepo — Fetch dataset from UCI repository

---

🧪 Model Evaluation Results

Metric	Value
R² Score	0.998
Mean Absolute Error (MAE)	0.0176
Mean Squared Error (MSE)	0.00098
Average Accuracy	96.67%
Mean Absolute Percentage Error (MAPE)	3.32%

✅ These results show the model predicts energy usage with extremely high precision.

---

📈 Visualizations

• Actual vs Predicted Scatter Plot
• Residual Distribution
• Feature Importance Bar Chart

These plots help visualize how close predictions are to actual values and which features influence power usage the most.

---

🚀 How to Run

1. Clone this repository

   git clone https://github.com/your-username/your-repo-name.git
   

2. Install required dependencies

   pip install 
   

3. Open the notebook

   jupyter notebook Electric_Power_Prediction.ipynb
   

4. Run all cells sequentially to reproduce results.

🏁 Results Summary

The Random Forest model provides near-perfect predictions for household energy consumption. It can be extended for real-time power monitoring, energy efficiency analysis, or smart grid applications.

🌟 Acknowledgments

UCI Machine Learning Repository for the dataset

Scikit-learn Team for ML tools

Google Colab for providing a free compute environment