Vehicle Sales Forecasting in Maryland

Forecasting monthly new and used vehicle sales in Maryland using time series models and unsupervised machine learning.

📌 Project Overview

This project analyzes over 20 years of monthly vehicle sales data in Maryland to forecast future trends and uncover seasonal patterns. Time series models like SARIMAX and Prophet were used for forecasting, and K-Means clustering was applied to discover seasonal groupings within the sales data.

---

🎯 Objectives

• Analyze long-term trends and seasonal components in vehicle sales.
• Forecast future sales using time series models (SARIMAX and Prophet).
• Apply K-Means clustering to identify seasonal patterns and segment time periods.
• Validate model accuracy using performance metrics (R² Score, MAE).
• Support data-driven planning for dealerships and inventory managers.

---

🛠️ Techn Used

• Programming Language: Python
• Libraries & Tools:
  • Pandas, NumPy
  • Matplotlib, Seaborn
  • SARIMAX (statsmodels)
  • Prophet (Meta’s forecasting tool)
  • Scikit-learn (K-Means Clustering)

---

Key Results

• Achieved R² score of 0.85 in time series model evaluations.
• Captured seasonal peaks and dips in vehicle sales over 260+ months.
• Identified lag features and clustering patterns to enhance model precision.
• Created clear visualizations of sales trends and predictions.

Use Cases

This project can benefit:

• Auto dealerships aiming to align inventory with seasonal demand.
• State transportation and economic agencies for policy planning.
• Data analysts interested in time series modeling and clustering techniques.