Machine Learning-Powered Mobile App for Predicting Used Car Prices
Publication Type
Conference Paper

Buying and selling used cars is a common practice in all countries. When looking to sell a car, the seller decides the price of
their car after monitoring the prices of similar cars in the advertisements. When someone is looking to buy a car, they watch
advertisements for similar cars to get an idea of the price of the car they want to purchase. Despite the availability of blue
books that provide an estimate of automotive pricing, real market prices vary depending on demand and supply. In this
paper, we applied cutting-edge machine learning techniques to automate this process. The training data set is up-to-date
and it was collected from an active commercial website. We created semi-automated rule-based scripts to clean and
prepare the data for machine learning. Several machine learning algorithms were explored to generate an approximate
value for the car pricing, including Artificial Neural Network, Support Vector Machine, K-Nearest Neighbors, Random Forest,
and Gradient Boosted Decision Tree. To determine the features that most affect the price, extensive data analysis and
cleaning were undertaken. Our findings indicate that the testing accuracy is 90%. Finally, a mobile application was created
that provides an estimated price of a given car’s properties, guiding a user in determining the price of their car. The
complete code and the data used to obtain these results can be accessed on GitHub at [1][19].

Conference Title
3rd International Conference on Big-data Service and Intelligent Computation (BDSIC 2021)
Conference Country
Conference Date
Nov. 9, 2021 - Nov. 11, 2021
Conference Sponsor
Additional Info
Conference Website