Machine-Learned Image Selection for Conversion and Relevance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image presentation methods in delivery applications often display irrelevant or low-quality images, leading to user confusion and wasted computing resources, as they fail to optimize image selection based on user preferences and conversion likelihood.
Innovation Solution
A central server computer employs a scoring algorithm that combines a conversion component and an uncertainty component to dynamically select and rotate images, using a multi-arm bandit algorithm to balance exploitation and exploration, ensuring images displayed are relevant and engaging.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional image presentation methods are used, then implementation is simple, but image relevance and conversion rate deteriorate
Solution Approach 1:
The system changes parameters by introducing a scoring mechanism that evaluates images based on multiple factors including conversion probability and uncertainty. This transforms the image selection from a static process to a dynamic one where images are scored and selected based on calculated metrics rather than simple rotation or manual selection.
Solution Approach 2:
The patent replaces mechanical/manual image selection methods with an automated machine learning-based scoring system. The multi-arm bandit algorithm automatically determines which image to display by calculating scores based on conversion probability and uncertainty, eliminating the need for manual intervention or simple rotational algorithms.
2Adaptability or versatility
If image rotation is increased to maintain user interest, then engagement may improve, but user confusion and frustration increase
Solution Approach 1:
The system implements dynamic image selection where the displayed image changes based on real-time scoring. Instead of fixed rotation schedules, the system adapts image presentation based on calculated conversion probabilities and uncertainty metrics, allowing the image selection to dynamically respond to performance data while maintaining user understanding through relevance.
Solution Approach 2:
The patent incorporates feedback loops where image performance is continuously monitored and used to update scoring models. The multi-arm bandit algorithm uses feedback from conversion events to adjust which images are selected, ensuring that image variety is provided only when it leads to better conversion outcomes rather than causing confusion.
3Reliability
If machine learning scoring algorithm is implemented, then image selection quality improves, but computational resources increase
Solution Approach 1:
The system applies partial action by using a simplified multi-arm bandit approach rather than full-scale complex machine learning models. The scoring algorithm focuses on key metrics (conversion probability and uncertainty) rather than analyzing all possible image attributes, providing good enough selection accuracy with reduced computational overhead compared to exhaustive ML approaches.
Solution Approach 2:
The patent extracts only the essential components needed for effective image selection - conversion probability and uncertainty metrics - from a potentially complex machine learning system. By focusing on these two key factors and using a multi-arm bandit framework, the system achieves reliable image selection without the excessive computational resources required by comprehensive ML models.
Data Source
AI summary
A method is disclosed. The method includes receiving, by a server computer, a plurality of images associated with one or more service providers. The server computer then receives an inquiry request, and determines an image of the plurality of images. The image is selected in response to a composite score based on a scoring algorithm scoring each image in the plurality of images. The scoring algorithm comprises a conversion component and an uncertainty component. The server computer provides an inquiry response comprising the image to the end user device operated by the end user.