Modeling Online Auctions / Edition 1by Wolfgang Jank, Galit Shmueli
Pub. Date: 07/13/2010
The new data challenges associated with online auctions motivate the need for clever statistical ideas and new statistical innovation in order to gain knowledge about bidders, sellers, prices, and a host of other questions of interest. In this book, the authors draw upon their experience of working with online auction data and introduce the reader to… See more details below
The new data challenges associated with online auctions motivate the need for clever statistical ideas and new statistical innovation in order to gain knowledge about bidders, sellers, prices, and a host of other questions of interest. In this book, the authors draw upon their experience of working with online auction data and introduce the reader to state-of-the-art statistical methodology for extracting new knowledge from online auction data. Rather than approach the topic from the traditional game-theoretic route, the authors treat the online auction mechanism as a new type of data generator, as well as use statistical and data mining methods to collect, explore, model, and forecast data that arises from online auction databases. Every effort is made to embellish cross-disciplinary fertilization between statistics, data mining, marketing, information systems, and economics and related fields.
Table of Contents
1.1 Online Auctions and Electronic Commerce.
1.2 Online Auctions and Statistical Challenges.
1.3 A Statistical Approach to Online Auction Research.
1.4 The Structure of this Book.
1.5 Data and Code Availability.
2 Obtaining Online Auction Data.
2.1 Collecting Data from the Web.
2.2 Web Data Collection and Statistical Sampling.
3 Exploring Online Auction Data.
3.1 Bid Histories: Bids versus "Current Price" Values.
3.2 Integrating Bid History Data With Cross-Sectional Auction Information.
3.3 Visualizing Concurrent Auctions.
3.4 Exploring Price Evolution and Price Dynamics.
3.5 Combining Price Curves with Auction Information via Interactive Visualization.
3.6 Exploring Hierarchical Information.
3.7 Exploring Price Dynamics via Curve Clustering.
3.8 Exploring Distributional Assumptions.
3.9 Exploring Online Auctions: Future Research Directions.
4 Modeling Online Auction Data.
4.1 Modeling Basics (Representing the Price Process).
4.2 Modeling The Relation Between Price Dynamics and Auction Information.
4.3 Modeling Auction Competition.
4.4 Modeling Bid and Bidder Arrivals.
4.5 Modeling Auction Networks.
5 Forecasting Online Auctions.
5.1 Forecasting Individual Auctions.
5.2 Forecasting Competing Auctions.
5.3 Automated Bidding Decisions.
and post it to your social network
Most Helpful Customer Reviews
See all customer reviews >
Kisses you back and closses my eyes putting my hands on your shoulders
Ok ..smirks...he grabs her roughly and presses his lips into hers kissing her hard (how about we are both already na.ked
This book can be read as a sequel to an earlier text, Statistical Methods in e-Commerce Research (Statistics in Practice). That looked at various types of websites, like Amazon and Wikipedia, while the current offering specialises to the key case of studying auction websites. Of these, eBay dominates the discussion, simply because it is the largest such website on the Internet. In a way, the exposition is simpler than it might have been just a few years ago, when Amazon and Yahoo also ran auctions. They were keying off eBay's success and hoped to take some of that business. But their efforts came to nought, so that now if you study online auctions, it is really only eBay and a handful of much smaller entities like uBid. The screen captures in chapter 2 of typical web pages from an auction [on eBay] shows the complex spaghetti-like source code. Note that if you do decide to screen scrape, then this is brittle, since if the website makes just minor changes in the format of their pages, extensive changes to your parsing might be necessary, to extract the same information. But as the authors make clear, screen scraping has the advantage of being free. An alternative is to use a Web Service, if that is offered by the website. Much more efficient and robust. But not all websites have this, and those that do could require payment. Plus the information offered by their Web Service might not include data that you need. Chapter 3 tackles the problem of how to simulate data using continuous distributions, when actual auction data often looks like a mixture of continuous and spiky inputs, where the latter are bids of high frequency, that stand significantly above the rest of the bid distribution. Chapter 4 discusses various models - exponential, log, logistic and inverse logistic, that can be used to model a given auction. But the problem is that across a set of auctions, even for instances of the same item being offered, all such models might be observed. Where a specific auction could be best fitted by a log, say, while another auction looks like a logistic. The authors suggest a "Beta()" function that has only 2 parameters. This turns out to be easy to compute, and, depending on the choices of parameter values, can replicate each of the 4 earlier models. Just as importantly, the fitting of a Beta to a given auction can be automated, which gets around an earlier problem of having to make a manual choice between one of the earlier models. Perhaps as interestingly, the chapter goes further, into studying what the book calls the spatial similarity between auctions. The spatial refers to auctions where items differ slightly. For example, a gaming computer that has different colours, and different disk sizes and different memory sizes installed. This is a multidimensional feature space, where the features might differ continuously [like memory] or discretely [like colour]. It reflects the well known attraction of eBay where if you search for a popular item, you can find hundreds [or even thousands] that differ in parameters like these. Modelling the behaviour of bidders when confronted by a surfeit of choices would be good. The authors show a way to tackle how to define a metric in the feature space.