Maaz Hussain

Data Science Analysis of the Used Car Market

Methodology

The project follows a structured workflow:

  1. Data Preparation:
    • Converted data types to their proper formats.
    • Addressed missing values, duplicates, and outliers.
    • Engineered features such as car age and mileage to enhance predictive power.
  2. Exploratory Data Analysis (EDA):
    • Statistical Measures: Calculated measures of central tendency (mean, median) and spread (standard deviation, range).
    • Distribution Analysis: Assessed skewness and kurtosis to understand data symmetry.
    • Visualization Tools: Plots and heatmaps were used to identify relationships between numerical and categorical variables.
  1. Insights from Data Visualization:
    • Car Age vs Selling Price: Newer cars fetch higher prices, indicating depreciation over time.
    • Mileage vs Selling Price: Vehicles with better mileage command higher prices, reflecting buyer preference for fuel efficiency.
    • Fuel Type vs Selling Price: Diesel cars are valued higher due to their superior mileage and efficiency.
    • Seller Type vs Selling Price: Cars sold by certified dealers often have higher prices, reflecting trust in quality.
    • Transmission Type vs Selling Price: Automatic vehicles command premium prices, likely due to higher manufacturing costs and demand.
  1. Correlation Analysis:
    • Selling price strongly correlates with car_age, max_power (bhp), and torque (Nm), highlighting these as key factors in price prediction.
  1. Machine Learning Model Development:
    • Applied linear regression to predict selling prices using significant features such as car age, engine capacity, max power, mileage, and fuel type.
    • Used MinMaxScaler for feature scaling and cross-validation to ensure model reliability and prevent overfitting.
    • Explored feature importance and engineered polynomial features for enhanced model performance.
  1. Model Evaluation:
    • Cross-validation ensured robust performance across folds.
    • Comparative analysis of predictions on training and test datasets validated model generalizability.

Key Insights

  1. Feature Importance:
    • Key predictors for selling price include car_age, max_power (bhp), engine_capacity (cc), and torque (Nm).
    • Categorical factors like fuel type and transmission type also significantly impact pricing.
  2. Trends:
    • Depreciation: Selling price decreases as car age increases.
    • Efficiency Matters: Buyers prefer cars with higher mileage and better fuel types, such as diesel.
    • Transmission Preference: Automatic vehicles are priced higher than manual ones.
  3. Market Dynamics:
    • First-owner cars are preferred, reflecting better condition and higher trustworthiness.
    • Certified sellers (trustmark dealers) command higher prices compared to individual sellers.

Conclusion