Predicting Used Car Prices in Metro Manila Using Artificial Neural Networks on Web-Scraped Data

This study developed and assessed the effectiveness of an Artificial Neural Network (ANN) model using Python within the Spyder integrated development environment for predicting used car prices in Metro Manila, Philippines. The dataset was sourced from “cars.trovit.ph” scraped using a browser extension “Instant Data Scraper” and encompassed car attributes like make, body type, fuel type, engine type, year of release, and mileage. The ANN model, a Multilayer Feedforward Neural Network (MFFNN) with Back Propagation Algorithm, underwent meticulous hyperparameter tuning, resulting in significant improvements. These enhancements encompassed notable reductions in Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE), alongside elevated R-squared (R2) and Adjusted R-squared values. Evaluation involved rigorous assessments, including K-Fold cross-validation and holdout validation techniques. The achieved results showcased the model’s proficiency, with performance metrics as follows: MSE = 31,905,395,405.08, RMSE=179,402.63, MAE =125,546.52, R2 Score = 0.7031, and Adjusted R2 Score =0.7031.