Flight Ad Ranking Using Pretrained Models for Limited Bandwidth
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Targeted advertising on aircraft is difficult due to limited data processing power and communication bandwidth, making it challenging to deliver personalized advertisements to passengers effectively.
Innovation Solution
Implementing a computer-implemented method using machine learning algorithms, specifically a learn-to-rank algorithm, to train a model that ranks electronic advertisements based on flight details and passenger behavior data, enabling targeted advertisement selection and optimization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If targeted advertising is implemented on aircraft, then advertisement relevancy and effectiveness are improved, but data processing requirements and communication bandwidth increase
Solution Approach 1:
The system performs preliminary actions by collecting and processing passenger data before the flight, training machine learning models offline, and pre-generating personalized advertisement recommendations. This allows the aircraft system to operate with minimal onboard processing power while still delivering targeted advertisements during the flight.
Solution Approach 2:
The patent introduces an intermediary ground-based processing system that acts as a mediator between data collection and advertisement delivery. This ground-based system handles heavy data processing and model training, then transmits only essential model parameters and recommendations to the aircraft, reducing onboard computational requirements.
2Productivity
If personalized advertisements are delivered to each passenger, then conversion rates increase, but communication bandwidth requirements increase
Solution Approach 1:
The system extracts only the essential personalized advertisement recommendations and model parameters from the complete data processing pipeline, transmitting only this condensed information to the aircraft. This extraction approach delivers personalized ads to each passenger while minimizing communication bandwidth consumption by excluding redundant data.
3Measurement precision
If machine learning models are trained onboard, then advertisement targeting accuracy is improved, but device complexity and processing requirements increase
Solution Approach 1:
The system performs the complex machine learning model training in advance on ground-based servers before the flight. During the flight, the pre-trained models are deployed with minimal processing requirements, maintaining high targeting accuracy while avoiding the complexity of real-time onboard training.
Solution Approach 2:
The patent replaces the mechanical system of onboard training with a remote ground-based training system. This substitution transfers the computational burden from the aircraft's limited onboard processors to powerful ground-based servers, reducing device complexity while preserving model accuracy.
Data Source
AI summary
The present disclosure relates to ranking electronic advertisements using one or more machine learning algorithms for a targeted audience associated with an aircraft flight. For example, one or more embodiments described herein include a computer-implemented method comprising executing a learn-to-rank algorithm to train a machine learning model on a training dataset that includes electronic advertisements with associated scores characterizing a relevancy between the electronic advertisements and a defined query. The computer-implemented method can also comprise applying the trained machine learning model to rank a set of electronic advertisements based on a feature vector characterizing input data that includes flight details of an aircraft.