A Study Of Online Auction Processes Using Functional Data Analysis

Online auctions have become popular in E-commerce in recent years. One of the big players in this industry, eBay, recorded net revenue of $10.2 billion in 2020, which was up about 19% of the net revenue for 2019. The purpose of this study is to characterize and understand online auction processes using functional data analysis. The data set we used is a 7-day auction process of a palm pilot m515 on eBay. The method used involves using functional principal component analysis to split each auction into curves made up of an overall fixed mean function, a zero-mean random process with functional principal components, and a zero-mean random variable representing measurement errors. Functional clustering analysis was also performed on the auctions data using three different methods. To know the optimal number of clusters, three internal validity measures were used. After extensive clustering comparisons, the K-means method was identified as the best clustering method which gives us the most information about bids during the auction processes for this specific dataset.