A Comprehensive Exploration of Machine Learning Models for Predicting Online Auction Prices

Abstract

The transformative impact of traditional commerce by online marketplaces is exemplified through eBay, a global platform that facilitates diverse transactions via auctions. In this research, the dynamics of eBay auctions, crucial for buyers, sellers, and researchers, are delved into. The central inquiry revolves around the key factors shaping auction outcomes, examining bid behaviors and types. The study leverages a robust dataset from eBay, meticulously curated to encompass auction identifiers, bid details, pricing information, auction types, and temporal aspects. A comprehensive approach involves data preprocessing, ensuring reliability by addressing missing values and outliers. Rigorous exploration and validation validate the dataset's integrity. Machine Learning Techniques, including MLP, SVR, Linear Regression, Extra Trees, and Gradient Boosting, form the analytical backbone. Model evaluation reveals top-performing candidates, such as MLP Regressor (0.8084), SVR (0.8210), and Linear Regression (0.8173), exhibiting superior accuracy and reliability. These models are identified for adoption in future work, emphasizing nuanced predictions in eBay auctions. This research contributes to understanding online auction dynamics, offering practical insights for eBay users and the broader e-commerce community. The models identified pave the way for enhanced predictive capabilities and continuous refinement in deciphering factors influencing auction outcomes.