ML Location Verification via Transaction Token Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to accurately determine whether a data operation occurred at the indicated geographic location, especially in scenarios where traditional contact methods are unavailable, such as during natural disasters or terrorist attacks, due to the lack of clear indication of the user's location in operation data.
Innovation Solution
A system utilizing a computing engine with a machine-learning algorithm that processes operation data by converting character fields into tokens, retrieving likelihood values and coefficients, and executing these to determine the probability of the operation occurring at the indicated location, thereby distinguishing between in-person and remote transactions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional means of contacting or locating a person are used, then the method is simple and widely applicable, but the method fails when cell service is down or the person has no functioning smartphone
Solution Approach 1:
The patent uses transaction data from third-party merchants as an intermediary to indirectly determine user location. Instead of directly accessing the user's device or communication channels, the system analyzes transaction records that contain merchant location information, effectively using the merchant's location data as a mediator to infer the user's position without requiring direct contact with the user's device.
Solution Approach 2:
The patent replaces the mechanical/physical system of direct device communication (smartphone, cell service) with a data analysis system that processes transaction records. Instead of relying on physical device functionality to determine location, the system substitutes a computational approach that analyzes patterns in transaction data to infer location, eliminating the need for functioning communication hardware.
2Measurement precision
If operation data is used to determine geographic location, then location information can be obtained without cell service, but the operation data may not indicate whether the operation actually occurred at the indicated location
Solution Approach 1:
The patent segments the location determination problem into multiple analytical components: (1) extracting merchant location from transaction data, (2) analyzing device sensor data (GPS, accelerometer, gyroscope) to determine device orientation and movement patterns, (3) comparing transaction timing with device state, and (4) synthesizing these segmented pieces of evidence to infer whether the operation occurred at the merchant location or remotely.
Solution Approach 2:
The system uses feedback from multiple data sources to continuously refine location determination accuracy. Device sensors provide feedback on device orientation and movement, transaction timing provides feedback on operational context, and pattern recognition algorithms use feedback from historical data to improve inference accuracy. This multi-layered feedback mechanism allows the system to compensate for the lack of direct location information in transaction data.
3Measurement precision
If a machine-learning algorithm is used to process operation data, then the accuracy of determining whether an operation occurred at the indicated location is improved, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary actions by pre-processing and organizing transaction data, device sensor data, and historical patterns before the actual location determination is needed. The system pre-establishes patterns of normal versus fraudulent transaction behaviors, pre-categorizes merchant locations, and pre-configures the machine learning model with relevant features. This preliminary preparation reduces the computational burden during real-time verification and simplifies the complexity of the overall system.
Data Source
AI summary
In one embodiment, the present disclosure includes a method for determining a location of a user. In one embodiment, the method includes receiving operation data including character fields and a first identifier. A query that includes the first identifier is sent to a data store. In response to the query, operation types and corresponding count values are received. Character fields from the operation data are converted into tokens. For each of the tokens, a likelihood value and a second identifier specifying a respective machine learning parameter is retrieved. A machine-learning algorithm is configured based on the count values, and the likelihood values and specified parameters are processed by the configured machine-learning algorithm to determine whether the operation data corresponds to the geographic location of a user.