Building Energy Consumption Prediction

Python, Machine Learning, Scikit-learn, Pandas, NumPy, EDA, Feature Engineering, Régression, GridSearchCV, Data Visualization

Prediction of Building Energy Consumption

Context

As part of a Data Science project, I developed a Machine Learning model to predict the energy consumption of non-residential buildings in the city of Seattle based on their structural characteristics.

The objective was to propose a solution capable of estimating the consumption of buildings for which no measurements were available, in order to support carbon emission reduction policies.

Objectives

• Perform an exploratory data analysis (EDA).
• Clean and prepare the data.
• Design new variables (Feature Engineering).
• Compare several regression models.
• Optimize the best model.
• Identify the factors with the most influence on predictions.

Architecture

Dataset Seattle Buildings
        │
        ▼
Exploratory Analysis (EDA)
        │
        ▼
Data Cleaning
        │
        ▼
Feature Engineering
        │
        ▼
Data Preparation
(Encoding, Scaling)
        │
        ▼
Comparison of several models
        │
        ▼
Optimization (GridSearchCV)
        │
        ▼
Predictions + Feature Importance

Technologies Used

• Python
• Pandas
• NumPy
• Matplotlib
• Scikit-learn
• Jupyter Notebook

Achievements

Exploratory Analysis (EDA)

• Study of data quality.
• Analysis of distributions.
• Identification of outliers.
• Analysis of correlations.
• Visualization of main building characteristics.

Data Preparation

• Handling missing values.
• Removal of inconsistent data.
• Encoding categorical variables.
• Normalization of numerical variables.
• Constitution of training and test sets.

Feature Engineering

Creation of new variables to improve model performance:

• temporal characteristics (year of construction, age)
• building uses
• structural variables
• category grouping
• reduction of data leakage risk

Modeling

Comparison of several supervised regression models:

• Linear Regression
• Random Forest Regressor
• Decision Tree Regressor
• Gradient Boosting Regressor

Model evaluation using cross-validation to select the best performing one.

Optimization

• Hyperparameter search with GridSearchCV
• Cross-validation
• Selection of the best model

Interpretation

Analysis of variable importance to identify the main factors influencing building energy consumption.

Demonstrated Skills

Data Science

• Exploratory Analysis (EDA)
• Data Cleaning
• Data Visualization
• Feature Engineering
• Outlier Detection

Machine Learning

• Supervised Regression
• Cross-validation
• Model Evaluation
• Hyperparameter Optimization
• Data Leakage Prevention

Python Development

• Data Manipulation
• Preparation Pipelines
• Jupyter Notebook
• Results Visualization

Business Analysis

• Understanding energy challenges
• Interpretation of explanatory variables
• Presentation of results to a business audience

Outcome

This project allowed me to implement the entire lifecycle of a Machine Learning project, from data analysis to the optimization and interpretation of a predictive model, while applying best practices for data preparation and performance evaluation.

Skills Acquired

• Python
• Pandas
• NumPy
• Scikit-learn
• Machine Learning
• Regression
• Data Science
• Feature Engineering
• Data Cleaning
• Data Visualization
• EDA
• GridSearchCV
• Cross Validation
• Random Forest
• Gradient Boosting
• Decision Trees
• Linear Regression
• Feature Importance
• Jupyter Notebook