Car CO2 Emissions: The Eco-Audit Challenge

K-12 Data Science: Multiple Regression, Regulatory Audits, and EV Disruption

Affiliate Disclosure: Some links below are affiliate links. If you purchase hardware or data tools through these links, this site earns a small commission at no additional cost to you, helping fund our K-12 STEM outreach programs.

Project Overview

Welcome to the Environmental Compliance Agency (ECA). An auto manufacturer is claiming their new fleet of vehicles has an ultra-low carbon footprint. Your team’s job is to audit their specifications using historical fleet data. We will fetch a live dataset directly from the web, visualize the distributions, overlay regulatory thresholds, and train a Linear Regression machine learning model to verify their emission claims.

Data Attribution: This dataset and core regression concept originate from the IBM Developer Skills Network / Cognitive Class Machine Learning with Python curriculum.

IBM Web Data (pd.read_csv) EDA Feature Extraction Engine, Cylinders, Fuel Scikit Linear Regression y = mx + b Eco-Audit Prediction


Step 1: Dynamic Data Ingestion

Modern data pipelines fetch data directly from web servers using code. We will use Pandas to download the IBM dataset.

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from sklearn import linear_model

# Ingest data directly from a public URL
url = "https://cf-courses-data.s3.us.cloud-object-storage.appdomain.cloud/IBMDeveloperSkillsNetwork-ML0101EN-SkillsNetwork/labs/Module%202/data/FuelConsumptionCo2.csv"
df = pd.read_csv(url)

# Select the features we want to investigate
cdf = df[['ENGINESIZE', 'CYLINDERS', 'FUELCONSUMPTION_COMB', 'CO2EMISSIONS']]
print(cdf.head())
   ENGINESIZE  CYLINDERS  FUELCONSUMPTION_COMB  CO2EMISSIONS
0         2.0          4                   8.5           196
1         2.4          4                   9.6           221
2         1.5          4                   5.9           136
3         3.5          6                  11.1           255
4         3.5          6                  10.6           244

Step 2: Exploratory Data Analysis (EDA) Histograms

Let’s view the distribution of our features to understand what kind of cars are in this dataset.

cdf.hist(figsize=(10, 6), color="#007acc", edgecolor="white")
plt.suptitle("Feature Distributions (Histogram Overview)", fontsize=14, fontweight='bold')
plt.tight_layout()
plt.subplots_adjust(top=0.9)
plt.show()

Step 3: Correlation Scatter Plots

Before building an AI model, we must prove a relationship exists. Let’s map Fuel Consumption vs. CO2 Emissions.

plt.figure(figsize=(8, 5))
plt.scatter(cdf.FUELCONSUMPTION_COMB, cdf.CO2EMISSIONS, color='#ff8c00', alpha=0.6, edgecolors='w')
plt.title("Fuel Consumption vs CO2 Emissions", fontsize=14, fontweight='bold', pad=12)
plt.xlabel("Combined Fuel Consumption (L/100km)", fontsize=12)
plt.ylabel("CO2 Emissions (g/km)", fontsize=12)
plt.grid(True, linestyle="--", alpha=0.5)
plt.tight_layout()
plt.show()

Step 4: The Regulatory Eco-Audit

The EPA has issued a strict emissions threshold of 250 g/km. Let’s plot Engine Size vs. CO2 Emissions and overlay the regulatory red line. How many vehicles are failing the audit?

plt.figure(figsize=(8, 5))
plt.scatter(cdf.ENGINESIZE, cdf.CO2EMISSIONS, color='#007acc', alpha=0.6, edgecolors='w', label="Vehicle Fleet")
plt.axhline(y=250, color='r', linestyle='-', linewidth=2, label="EPA Threshold (250 g/km)")
plt.title("Eco-Audit: Engine Size vs Regulatory Threshold", fontsize=14, fontweight='bold', pad=12)
plt.xlabel("Engine Size (Liters)", fontsize=12)
plt.ylabel("CO2 Emissions (g/km)", fontsize=12)
plt.legend()
plt.grid(True, linestyle="--", alpha=0.5)
plt.tight_layout()
plt.show()

Step 5: Training the Linear Regression Model

We split our data (80% for training, 20% for testing) to prevent the AI from memorizing the answers, then fit our regression model.

# 80/20 Train/Test Split
msk = np.random.rand(len(df)) < 0.8
train = cdf[msk]
test = cdf[~msk]

# Train the Simple Linear Regression Model
regr = linear_model.LinearRegression()
train_x = np.asanyarray(train[['ENGINESIZE']])
train_y = np.asanyarray(train[['CO2EMISSIONS']])
regr.fit(train_x, train_y)

print('Coefficient (Slope): ', regr.coef_[0][0])
print('Intercept: ', regr.intercept_[0])
Coefficient (Slope):  38.68356964337772
Intercept:  126.6935199003735

Step 6: Visualizing the Model & EV Disruption

The solid line represents our AI’s learning (y = mx + b). What happens if we release an Electric Vehicle (EV) with an Engine Size of 0.0L and 0 CO2 emissions? Does the model predict zero, or is there an intercept error?

plt.figure(figsize=(8, 5))
plt.scatter(train.ENGINESIZE, train.CO2EMISSIONS, color='#10b981', alpha=0.5, edgecolors='w', label="ICE Vehicles")
plt.plot(train_x, regr.coef_[0][0]*train_x + regr.intercept_[0], '-k', linewidth=2, label="Regression Line (AI Model)")

# EV Disruption Data Point
plt.scatter([0], [0], color='#ef4444', s=150, zorder=5, edgecolors='black', label="Zero-Emission EV (0L, 0g/km)")

plt.title("Machine Learning Fit & EV Disruption", fontsize=14, fontweight='bold', pad=12)
plt.xlabel("Engine Size (Liters)", fontsize=12)
plt.ylabel("CO2 Emissions (g/km)", fontsize=12)
plt.legend(loc="upper left")
plt.grid(True, linestyle="--", alpha=0.5)
plt.tight_layout()
plt.show()

Pedagogical Goals & Extension Ideas

  • Data Literacy: Ensure students understand the difference between a histogram (data spread) and a scatter plot (variable relationship).
  • The “What-If” Intercept Lesson: The final plot is a crucial talking point. The regr.intercept_ is around 125. Ask students: If an EV has a 0-liter engine, our math (y = mx + 125) says it emits 125g of CO2. Why is the AI wrong? This highlights the danger of out-of-distribution prediction.
  • Challenge Activity: Have advanced students rewrite the scikit-learn code to perform Multiple Linear Regression by feeding it train[['ENGINESIZE', 'CYLINDERS', 'FUELCONSUMPTION_COMB']] to see if the error improves compared to using just engine size.