blocPower

BlocPower Demo

This demo involves modeling energy usage for buildings in NYC.

Highlights

Data Import and Exploration

• Create Spark DataFrames from imported CSV files
• Run some data exploration commands to inspect the data
• Define functions to clean and prepare the data for modeling

Model Creation

• Create a model using sklearn
• Visualize the model accuracy
• Detect buildings that consume energy inefficiently

Project-Specific Information

Tools Used

• Python, through Jupyter Notebooks
• Spark DataFrames
• Scikit Learn for model creation

Model Information

Linear Regression

• Response variable is energy usage in kWh
• Features include age of the building, square feet, number of stories, total plugged equipment, etc.

Unsupervised Learning

• Perform PCA and Clustering
• Metrics used: plugged-in equipment, air conditioning, domestic gas, heating gas
• Visualize clusters using two out of four dimensions and K-means to determine center locations

Classification using Logistic Regression

• Model used to identify inefficient buildings
• Visualize accuracy using a confusion matrix

Relevant Files

Notebooks

• BlocPower.ipynb

Data Assets

• BlocPower_T.csv
• CDD-HDD-Features.csv
• HDD-Features.csv