/predicting_car_prices

Use k-nearest neighbors algorithm to predict car price based on physical characteristics

Primary LanguageJupyter Notebook

Predicting Car Prices

by Nicholas Archambault

This project predicts car prices based on a variety of technical characteristics, including engine size, fuel efficiency, and horsepower. Using the k-nearest neighbors algorithm with varying k-values and feature combinations allows for visualization and understanding of what makes an effective and accurate KNN model.

Goals

  1. Prepare dataset for machine learning applications with rigorous cleaning process.
  2. Initiate and gauge the comparative results of univariate and multivariate k-nearest neighbors models in order to understand which features most accurately predict car price.
  3. Modify k-values and tune hyperparameters to achieve more accurate predictions.
  4. Plot results to visualize the benefits and compromises of k-nearest neighbors testing.

Output

Analysis of the success of univariate, multivariate, and hyperparameter-tuned k-nearest neighbors models and how mean squared error changes with different model attributes.