This Jupyter Notebook is a practice/learning project that applies Linear Regression and Decision Tree Regression to a clean, small dataset of car specifications. The goal is to predict CO₂ emissions based on engine volume and vehicle weight.

• Source file: DATA1.csv

• Size: 36 entries, 5 columns

• Columns:

  Column	Type	Description
  Car	object	Car brand (e.g., Mercedes, Ford, BMW)
  Model	object	Car model name
  Volume	int64	Engine displacement (cc)
  Weight	int64	Vehicle weight (kg)
  CO2	int64	CO₂ emissions (target variable)

• Missing values: None

• Reads the CSV file using pandas
• Displays shape, column types, and missing value counts
• Shows value counts for each column

• Bar charts for car brand and model distributions
• Histogram of CO₂ emissions
• Boxplot for Weight and Volume
• Descriptive statistics table (max, min, quartiles, IQR, mean) for Weight, Volume, and CO₂
• Scatter plots: Volume vs. Weight, Volume vs. Car brand, Weight vs. CO₂
• Correlation matrix (heatmap) for Volume, Weight, and CO₂

• Features: Volume, Weight
• Target: CO2
• Uses sklearn.linear_model.LinearRegression
• Prints the regression equation coefficients and intercept
• Makes a sample prediction for a car with Volume=1300, Weight=2300

• Features: Volume, Weight
• Target: CO2
• Uses sklearn.tree.DecisionTreeRegressor with max_leaf_nodes=20
• Splits data into train/validation sets using train_test_split
• Evaluates with Mean Absolute Error (MAE)

• Uses sklearn.ensemble.RandomForestRegressor
• Same train/val split as the decision tree
• Also evaluated with MAE for comparison

• Correlation:

  • Volume ↔ Weight: ~0.75 (strong positive)
  • Volume ↔ CO₂: ~0.59 (moderate positive)
  • Weight ↔ CO₂: ~0.55 (moderate positive)

• Linear Regression Equation (approximate):

  CO2 ≈ 0.0078 × Volume + 0.0076 × Weight + 79.69
  

• MAE (Decision Tree & Random Forest): ~5.38 on validation set

pandas
numpy
matplotlib
seaborn
scikit-learn

• This notebook is for learning/practice purposes on a clean, small dataset.
• The dataset path is hardcoded (D:/chenghan/...) and will need to be updated to run locally.
• No formal train/test split is used for the Linear Regression model (fit on full dataset).