Cascading ML Floor Price Prediction for Auction Revenue
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current online advertisement auction systems face challenges in accurately predicting a dynamic floor price, which affects the revenue maximization for publishers in second price auctions, as existing methods struggle to balance between predicting a floor value close to the highest bid without exceeding it.
Innovation Solution
A cascading classification strategy utilizing machine learning techniques, where each node corresponds to a trained machine learning model, predicts a dynamic floor price by filtering and classifying bids through multiple layers to determine the optimal floor value that maximizes revenue, ensuring the winning advertiser pays the higher of the second highest bid or the predicted floor price.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If a dynamic floor price prediction system is implemented to maximize publisher revenue, then the revenue for publishers is improved, but the complexity of the auction system increases
Solution Approach 1:
The patent segments the auction system into multiple independent components: a floor price prediction module that uses machine learning to predict minimum acceptable prices, a bid evaluation module that compares bids against the predicted floor price, and a winner selection module. This segmentation allows the complex revenue maximization task to be handled by specialized sub-components rather than a monolithic system, improving revenue while managing complexity through modular design.
Solution Approach 2:
The system performs preliminary action by predicting the floor price before the auction actually occurs. The machine learning model analyzes historical bid data, user behavior patterns, and advertisement characteristics in advance to establish a data-driven minimum price threshold. This preliminary price setting guides the subsequent bidding process, ensuring revenue optimization without requiring complex real-time adjustments during the auction.
2Measurement precision
If machine learning models are used to predict floor prices, then the accuracy of price prediction is improved, but the computational time and resources increase
Solution Approach 1:
The machine learning models are trained offline on historical auction data, user behavior patterns, and advertisement characteristics before deployment. This preliminary training phase allows the models to learn complex patterns and relationships in advance. During actual auction execution, the pre-trained models can quickly predict floor prices with high accuracy without requiring extensive computational resources in real-time, thus resolving the trade-off between prediction accuracy and computational time.
Data Source
AI summary
Systems, devices, and methods are disclosed for predicting a dynamic floor price for increasing cleared revenue cleared after a winning bid is determined in an online bid auction. The dynamic floor price is predicted from a cascading classifier strategy implemented through a series of cascading machine learning based classifier models that have been trained.