Car Price Prediction: A Comparative Study of Machine Learning and Deep Learning Approaches

Accurately predicting used car prices is a complex challenge in the automotive sector, influenced by factors like brand, age, engine capacity, and mileage. This study develops predictive models using machine learning and deep learning to address these complexities. A publicly available dataset is preprocessed by handling missing data, encoding categorical variables, and normalizing features. Exploratory Data Analysis (EDA) identifies key pricing factors, trends, and patterns. Predictive models, including Linear Regression, Random Forest, Gradient Boosting, LSTM, CNN, and DNN, are evaluated using R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, MAE, and RMSE. Results demonstrate that deep learning models, particularly LSTM, outperform traditional machine learning, achieving an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.67, by capturing complex variable interactions. LSTM's superior performance highlights its ability to learn intricate relationships within the data. While Random Forest and Gradient Boosting show competitive performance, their limitations in capturing non-linear patterns are evident. CNN also demonstrates strong performance within the deep learning category. The study highlights the importance of data preprocessing, feature selection, and hyperparameter tuning. These findings underscore the potential of combining machine learning and deep learning to enhance pricing transparency and reliability in the used car market, paving the way for future research to explore larger datasets and external economic factors.