Empirical studies on price determinants of online auctions with machine learning applications
MetadataShow full item record
Current technological developments have changed our trading habits and the importance of e-commerce in our lives has grown rapidly in the past decade. This new economic and technological environment generates massive, cheap, easily accessible and invaluable data. One of the important topics in electronic trade is the price estimation. Electronic trade takes place usually through two sales methods. The first is auctioning and the second is Buy-it-Now (BIN) sales. This dissertation concentrates on the determinants of online auction end prices in the smartphone markets. In this context, 444 auction and 676 BIN sales realized between March-July 2018 were analyzed with current Machine Learning (ML) algorithms. As a new contribution to the literature, vendors’ descriptions are analyzed with Natural Language Processing (NLP), prices of similar products are taken into account and the effect of information from the bids in the initial stage are investigated with image processing algorithms. The analyses show that vendor’s descriptions, prices of similar products, information from the bids in the initial stage, auction length, the day auction starts, product accessories, the number of visits to the vendor profile have positive effects on auction prices. On the other hand, sellers' reputation, especially negative reviews adversely affect auction prices.