Gaussian Process Bid Optimization for Non-Linear Auction Spaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current search engine optimization methods fail to account for non-linearity in auction spaces, do not consider profit over cost, and lack adaptability to changes in the auction environment, leading to suboptimal keyword targeting and positioning.
Innovation Solution
The use of a Gaussian Process Model combined with Thompson sampling reinforcement learning to predict target positions and bid strategies, which considers non-linear relationships and adapts to changing auction dynamics by analyzing historical data and adjusting bids based on profit margins.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If simple linear assumptions are made about auction space relationships, then the optimization model is easier to implement, but it fails to capture non-linear relationships leading to suboptimal results
Solution Approach 1:
The patent transforms the auction optimization problem by changing the parameter representation from simple linear bid scores to a Gaussian process model that captures non-linear relationships. The model uses kernel functions to represent complex interactions between bid amount, quality score, and position, enabling accurate modeling of non-linear auction dynamics while maintaining computational feasibility through Bayesian inference.
2Device complexity
If reinforcement learning is not used, then the system is simpler to maintain, but it adapts slowly to changes in auction space and seasonal effects
Solution Approach 1:
The patent implements a feedback mechanism where the Gaussian process model continuously learns from observed auction outcomes. The model updates its predictions based on actual click-through rates, conversion data, and position changes, allowing it to adapt to seasonal effects, competitor behavior changes, and auction space dynamics without requiring complex reinforcement learning architectures.
3Ease of operation
If bid optimization focuses only on maximizing click rate or impressions, then the optimization target is simpler to define, but it does not consider profit generated compared to bid cost
Solution Approach 1:
The patent changes the optimization parameter from simple click rate or impression maximization to profit-aware bid scoring. The Gaussian process model incorporates profit margins, conversion values, and bid costs into a comprehensive objective function, allowing the system to optimize for actual business value rather than just traffic metrics. This enables the model to determine when to bid aggressively on high-value keywords and when to conserve budget on lower-return opportunities.
Data Source
AI summary
Disclosed are methods, systems, and non-transitory computer-readable medium for targeting bid and position for a keyword. For instance, the method may include obtaining information about the keyword, the information about the keyword including observations of value with respect to position for the keyword. The method may further include applying a Gaussian Process Model on the observations to obtain a prediction function and associated uncertainties, the prediction function and the associated uncertainties relating positions to expected values; applying a Thompson sampling reinforcement learning model on the expected values and the positions to obtain a target position; and applying a bid model to the target position to obtain bid information for the keyword. The method may also include transmitting a bid message to a search engine, the bid message including the bid information.