Predicting the End Prices of Online Auctions

Abstract

The eBay marketplace offers millions of items for sale each day. Understanding and being able to better predict the outcome of these auctions is important to buyers and sellers alike for individual profit maximization. Online auctions have become one of the fastest growing modes of online commerce transactions. eBay has 94 million active members buying and selling goods at a staggering rate. These auctions are also producing large amount of data that can be utilized to provide services to the buyers and sellers, market research, machine learning algorithm to predict end-prices of auction items. We describe the features used, and several formulations of the price prediction problem. Using the PDA category from eBay, we show that our algorithms are extremely accurate and can result in a useful set of services for buyers and sellers in online marketplaces. Online auctions are one the most popular methods to buy and sell items on the internet. With more than 100 million active users globally (ars of Q4 2011), eBay is the world's largest online marketplace, where practically anyone can buy and sell practically anything. The total value of goods sold on eBay was $68.6 billion, more than $2,100 every second. This kind of volume produces huge amounts of data that can be utilized to provide services to the buyers and sellers, market research, and product development. In this analysis, we collect historical auction data from eBay and use machine learning algorithms to predict sales results of auction items. We describe the features used and formulations used for making predictions. The algorithms used can be relatively accurate and can result in a useful set of services for buyers and sellers. Online auctions allow users to sell and buy a variety of goods, and they are now one of the most important web services. Predicting final prices on online auctions is a hard task. However, there has been much pioneering work over the past ten years. In this project an evaluation of the effectiveness of our method is also described in the project report. For final price prediction, we find that for multiclass binary prediction decision trees models is the best for prediction. We provide a discussion on the result, as well as some insight about our particular data set and avenues for future exploration.