Transaction Risk Model Using Key-Dependent Coefficients
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional predictive models face difficulties in characterizing cross interactions between categorical variables and other continuous or binary variables, which hinders their ability to accurately assess the risk of data transactions, such as fraudulent credit card payments, as the risk function conditioned on categorical variables significantly differs.
Innovation Solution
A predictive model that associates categorical variables with keys, generating composite values using coefficients for binary and continuous variables, allowing for scoring of data transactions through a linear or non-linear logistic regression model, potentially forming part of a neural network or support vector machine, and utilizing cascading sub-models to enhance predictive power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional predictive models use standard variable inputs, then the model structure remains simple, but the model cannot accurately characterize cross interactions between categorical and continuous variables
Solution Approach 1:
The model segments the interaction between categorical and continuous variables by introducing key-value pairs where categorical variables are mapped to keys, and their interactions with continuous variables are modeled separately through key-dependent coefficients. This segmentation allows accurate characterization of cross interactions while maintaining a structured approach to model complexity.
Solution Approach 2:
The patent adds a new dimension to the model by introducing keys as an intermediate layer between categorical variables and continuous variables. Instead of directly modeling interactions in the original variable space, the model transforms the problem into a key-space where interactions are captured through key-dependent coefficients, effectively adding a dimensional transformation layer.
2Reliability
If the model includes multiple types of variables simultaneously, then the predictive power is enhanced, but the model requires discretization which loses information
Solution Approach 1:
The model changes the parameter representation by using key-dependent coefficients instead of fixed coefficients. Continuous variables interact with categorical variables through keys, where each key has its own set of coefficients for continuous variables. This parameter change allows the model to capture interactions without discretizing continuous variables, preserving their full information content.
3Measurement precision
If the model uses key-dependent coefficients for continuous variables, then cross interactions are accurately captured, but the number of parameters increases significantly
Solution Approach 1:
The model performs preliminary action by pre-computing and storing key-dependent coefficients during the training phase. Instead of computing complex interactions in real-time during scoring, the model pre-processes the data to establish key-coefficient mappings, which are then efficiently applied during transaction scoring. This preliminary action reduces the computational burden during deployment.
Solution Approach 2:
The model structure allows each key to self-manage its own coefficients for continuous variables. Each key independently stores and applies its specific coefficient set when encountered during scoring, eliminating the need for complex global parameter management. This self-service approach simplifies the handling of numerous parameters through a decentralized structure.
Data Source
AI summary
In one aspect, input data for a predictive model characterizing a level of risk for a data transaction is received that includes values for categorical variables and one or more of binary variables and continuous variables the predictive model. Thereafter, one or more of the categorical variables is associated with one of a plurality of keys. Each key having corresponding coefficients for at least a subset of the binary variables and the continuous variables and the coefficients being dependent on a value for the key. A composite value based on values for each of at least a subset of the binary variables and the continuous variables as calculated using the corresponding coefficients for each key can then be generated. Scoring of the data transaction using the binary variables, the continuous variables, and the composite variables can then be initiated by the predictive model. Related apparatus, systems, techniques and articles are also described.