A procedure using support vector data description and mutual information for end price assessment in online C2C auction

We propose a systematic procedure for assessing the end price of an item in a C-to-C auction site. These sites deal with used product and the product features vary substantially even within a single product category that makes price assessment difficult. Besides, the true market demand at a particular time, the effect of spurious bidding activities also contributes to price variation. We suggest removing outliers, selecting the right features and clustering the product data can increase the prediction accuracy. Using a multivariate dataset from eBay on Dell Laptops, we show, the prediction accuracy using back propagation neural network improves considerably when used in combination with the methods for (1) removing outliers using Support Vector Data Description, (2) selecting the right features using mutual information as a measure, and (3) clustering of the datasets.