Car Price Prediction: Autotrader Machine Learning Project

This repository contains a comprehensive end-to-end Machine Learning pipeline designed to predict the selling price of used cars. Using a dataset of Autotrader vehicle adverts, the project explores the relationship between vehicle attributes (mileage, age, brand) and their market value.

---

Technical Overview

The project is implemented as a modular Jupyter Notebook (AT_ML.ipynb) that follows the standard ML lifecycle:

• Data Understanding: Initial exploration of features including price, mileage, and registration codes.

• Feature Engineering:

• Derivation of vehicle_age from registration data.

• Logarithmic transformations on target (price) and mileage to normalize right-skewed distributions.

• Handling high-cardinality features via Target Encoding and low-cardinality via One-Hot Encoding.

• Data Integrity: Implementation of Scikit-Learn Pipelines to ensure strict separation between training and test data, preventing data leakage during scaling and imputation.

• Predictive Modeling: Comparison of four regression algorithms: Linear Regression, K-Nearest Neighbors, Decision Trees, and Random Forest.

• Hyperparameter Tuning: Utilization of GridSearchCV to optimize model parameters such as n_neighbors and max_depth.

---

Key Features & Methodology

Data Preprocessing & Cleaning

• Missing Value Imputation: Numerical values are filled using the median, while categorical features use the most frequent (mode) strategy.
• Outlier Management: Extreme values in numerical columns are clipped to the 1st and 99th percentiles based on training set distributions.
• Category Grouping: Rare categorical values (frequency < 1%) are consolidated into an "Other" category to reduce noise and improve model generalization.

Advanced Visualizations

The notebook generates several diagnostic plots to validate the data at each stage:

• Missingness Maps: Heatmaps to visualize the success of the imputation strategy.
• Correlation Heatmaps: Identifying multicollinearity between numeric features.
• Residual Analysis: Scatter plots of predicted vs. actual values and error distributions to diagnose model bias.

---

Model Performance Metrics

Models are evaluated on unseen test data using the following metrics:

• ** Score**: Proportion of variance explained by the model.
• Root Mean Squared Error (RMSE): Error magnitude in the original currency (£) scale.
• Mean Absolute Percentage Error (MAPE): Relative prediction accuracy.

---

Requirements

To run this notebook, you will need the following Python libraries:

• pandas
• numpy
• scikit-learn
• matplotlib
• seaborn