Automobile Price Prediction

Machine learning regression model for predicting 1985 import vehicle prices. Achieves 91.7% test R-squared and 89.4% cross-validation R-squared with Lasso regression.

---

Overview

Predicts automobile prices from 25 vehicle specification features (brand, engine, body design, performance). Addresses data leakage, missing values (18%), multicollinearity (VIF > 16,000), outliers (10% of samples), and high-cardinality categoricals.

Final Model: Lasso Regression (alpha=10.0)

• Test: R-squared = 0.917, RMSE = $1,987
• Cross-validation: R-squared = 0.894 +/- 0.027
• Features: 42 (6 PCA + 36 categorical), 29 non-zero coefficients
• Overfitting gap: 3.3%

See Model_Comparison_Report.md for why Lasso was selected over XGBoost despite 16% higher test error.

---

Quick Start

Installation

pip install -r requirements.txt

Load Trained Model

import joblib
model = joblib.load('models/final_lasso_model.joblib')
predictions = model.predict(X_new)  # Requires 42 engineered features

Run Full Pipeline

Open notebooks/auto-price-prediction.ipynb for complete data preparation, modeling, and evaluation workflow.

Dataset

Property	Details
Source	1985 Auto Imports Database (UCI ML Repository)
Samples	200 vehicles (205 original, 5 removed)
Features	26 attributes (25 predictors + price)
Split	158 train / 40 test (stratified)

See data/raw/dataset-info.txt for full metadata.

Project Structure

Core directories:

• data/raw/ - Original dataset (auto_imports.csv)
• data/processed/train-test/ - Train/test split
• notebooks/ - Full analysis pipeline (auto-price-prediction.ipynb)
• src/ - Reusable modules (utils, statistical analysis, model evaluation)
• models/ - Trained model artifacts (final_lasso_model.joblib)
• reports/ - Detailed analysis reports

Working with the Notebook

Import pattern used: The notebook imports functions from src/ modules using:

from src.utils import memory_usage, dataframe_memory_usage
from src.statistical_analysis import normality_test_with_skew_kurt, spearman_correlation_with_target
from src.model_evaluation import evaluate_regression_model, hyperparameter_tuning

Running analysis: The notebook contains the full ML pipeline. Execute cells sequentially for:

1. Data loading and cleaning
2. Statistical analysis (normality tests, correlation)
3. Feature engineering (PCA on numerical features)
4. Model comparison (10 algorithms tested)
5. Hyperparameter tuning (5-fold GridSearchCV)
6. Final model selection and persistence

Model Training Workflow

Base model evaluation:

from src.model_evaluation import evaluate_regression_model

metrics = evaluate_regression_model(model, X_train, y_train, X_test, y_test)
# Returns: MAE, MSE, RMSE, R², Adjusted R², MSLE, MAPE, CV R², Training R², Overfit, Training Time

Hyperparameter tuning:

from src.model_evaluation import hyperparameter_tuning

best_models, best_params, times = hyperparameter_tuning(
    models={'Lasso': Lasso()},
    param_grids={'Lasso': {'alpha': [0.1, 0.5, 1, 5, 10, 50, 100, 500, 1000]}},
    X_train=X_train,
    y_train=y_train,
    scoring_metric='neg_mean_squared_error',
    cv_folds=5
)

Statistical Analysis Functions

Normality testing:

from src.statistical_analysis import normality_test_with_skew_kurt

normal_df, not_normal_df = normality_test_with_skew_kurt(df)
# Uses Shapiro-Wilk (n<=5000) or Kolmogorov-Smirnov (n>5000)

Multicollinearity detection:

from src.statistical_analysis import calculate_vif

vif_data, high_vif_features = calculate_vif(data, exclude_target='TARGET', multicollinearity_threshold=5.0)
# Returns VIF scores and features exceeding threshold

Spearman correlation:

from src.statistical_analysis import spearman_correlation_with_target

corr_data = spearman_correlation_with_target(
    data,
    non_normal_cols=['col1', 'col2'],
    target_col='TARGET',
    plot=True,
    table=True
)

Model Persistence

Loading the final model:

import joblib
model = joblib.load('models/final_lasso_model.joblib')
predictions = model.predict(X_new)

Key Design Decisions

Model selection criteria (weighted):

1. Generalization (CV R² stability) - 40%
2. Accuracy (Test RMSE/R²) - 30%
3. Stability (overfitting gap) - 20%
4. Efficiency (training time, interpretability) - 10%

Why Lasso over XGBoost:

• XGBoost achieved lower test RMSE (1,663 vs 1,987) but showed 8.3-point CV-test gap vs Lasso's 2.3-point gap
• Lasso provides interpretability (29 sparse coefficients) vs XGBoost black box
• Training: 11.5x faster (0.014s vs 0.161s)
• Inference: 600x faster operations
• Trade-off: Accept 2.5% higher error for 4.1 points better CV R² and full transparency

Reports

Detailed analysis in reports/:

• Complete_Data_Analysis_Report.md - Full methodology, statistical analysis, and results
• Model_Comparison_Report.md - Model selection rationale and performance comparison
• Challenges_Report.md - Technical challenges and solutions
• GALLERY.md - Visualizations

Development

Code Quality

# Format code
black .
isort .

# Run pre-commit hooks
pre-commit run --all-files

Pre-commit Hooks

• black (88-char lines)
• isort (black-compatible)
• nbqa-black (notebooks)
• Validation (YAML, JSON, trailing whitespace)

---

• MIT License - Copyright (c) 2025 Dhanesh B. B.
• GitHub: https://github.com/dhaneshbb