A reserve price optimization (RPO) system estimates bids of individual bidders for an auction based on bidding information specific to the individual bidder. Auction data and bidder specific information are input to a model (e.g., neural network) that determines an estimated bid. Each of the estimate bids for each of the individual bidders of an auction are used as inputs in determination of an auction floor estimate optimization for a current auction. The auction floor estimate can be discounted by a safety bias to decreases likelihood of the auction floor estimate surpassing an actual top bid of the bidders.
Greenberg, Don, "Auction Floor Optimization", Technical Disclosure Commons, (April 05, 2017)