Suppose I participate in online auctions. I submit bids based on the features of the item being sold, but I only got feedback when I win. If I lose, I would know nothing about who won, and how much was the winning price. I can only assume that you bidded too low.

I would like to use the data of my bidding history, both wins and losses, to predict the winning price and it's winning chance in percentage (the prediction interval). What machine learning problem is this?

I tried building a regression model based on item features using only the winning bid histories, and that worked to some degree. However, I imagine that will be biased, as the losing bid information is not incorporated. For example, if I have the following bid histories: (3, lose), (4, lose), (5, win), (6, win), (7, win), (8, win)

Using only the winning bid for regression and minimizing for MAE, I would get a price close to 6.5. However, if looking at the full bid history, the actual winning price might be close to 5. So using only the winning price is likely to over predict.

The use case is for real time online advertisement bidding if it's relevant.

  • $\begingroup$ Not sure if it is the best idea, but what did you try using the Bayes formula to estimate the probability that the price is above the given value? What would be the mean in this case? $\endgroup$ – kate-melnykova Dec 5 '20 at 4:54
  • $\begingroup$ No I haven't. Would you mind elaborate a bit more on how to use Bayes formula to estimate the probability of the winning price being above any given price? $\endgroup$ – Vance Dec 7 '20 at 0:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.