Machine Learning Classifier for Real-Time Traveler Segmentation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current online advertising systems face challenges in making real-time ad selection and bidding decisions due to insufficient user information available via ad exchanges, which limits the ability to utilize rich customer data from advertisers' databases, leading to performance issues and privacy concerns.

Innovation Solution

A method using machine learning classifiers to classify unidentified online users into traveler categories based on limited online information by accessing offline data stores, computing feature vectors, and training models to make rapid classification decisions, enabling targeted ad selection and bidding within milliseconds.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If additional cumulative information is gathered and stored about unidentified online users using browser cookies to enable matching with advertiser databases, then user identification accuracy improves, but data privacy issues and performance issues arise

Engineering Contradiction:

Improveuser identification accuracyVSAvoiddata privacy issues

Core Design Contradiction:

Measurement precisionVSObject-affected harmful factors

Solution Approach 1:

The patent introduces a machine learning model as an intermediary that processes online user information and offline customer data without directly exposing or storing sensitive personal information. The model learns patterns from offline data and applies them to online users through feature vectors, enabling identification without direct data linkage that would raise privacy concerns.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent extracts only the necessary features from offline customer data that can be represented in online contexts (e.g., travel preferences, booking behaviors) without extracting or storing sensitive personal identifiers. This extraction approach enables matching while minimizing privacy exposure by removing directly identifiable information.

Inventive Principle:

Principle #2Taking out (Extraction)

2Measurement precision

Engineering Contradiction:

Improveuser identification accuracyVSAvoidsystem performance

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent extracts only the essential features from offline customer data that are relevant to online user behavior (e.g., travel frequency, preferences, booking patterns). This selective extraction reduces the data volume significantly while maintaining the ability to perform accurate matching, thereby improving system performance.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms complex offline customer data into simplified feature vectors that capture essential user characteristics in a compact format. This parameter transformation reduces computational complexity and enables faster processing while maintaining identification accuracy.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If matching logic is made more comprehensive to cover various ways of combining information for determination, then user identification accuracy improves, but device complexity increases

Engineering Contradiction:

Improveuser identification accuracyVSAvoidmatching logic complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex mechanical matching logic with a machine learning model that has been trained offline. The model automatically learns the appropriate ways to combine and evaluate features without requiring explicit programming of complex matching rules, thereby reducing system complexity while maintaining or improving accuracy.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

4Speed

If real-time bidding decisions are made with limited user information available via ad exchanges, then processing speed improves, but user identification accuracy deteriorates

Engineering Contradiction:

Improvebidding decision speedVSAvoiduser identification accuracy

Core Design Contradiction:

SpeedVSMeasurement precision

Solution Approach 1:

The patent performs preliminary training of the machine learning model using offline customer data before real-time bidding decisions are made. This preliminary action enables the model to learn from comprehensive data offline, so that during real-time bidding with limited information, the model can make accurate predictions based on the pre-learned patterns without requiring additional processing time.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11120480B2Systems and methods for real-time online traveler segmentation using machine learning

Publication Date: 2021.09.14 AMADEUS SAS

AI summary

Methods and computing apparatus for real-time online traveler segmentation. A machine learning classifier may be trained using computed feature vectors and associated tags corresponding with records in a training set. A machine learning classifier receives a feature vector comprising values of the plurality of features corresponding with an unidentified user in an online context. The machine learning classifier may determine an estimate of whether the unidentified user is a member or a non-member of a predetermined traveler category.