Machine Learning Entity Segmentation with Distribution-Based Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing entity population segmentation techniques, such as mixed integer programming (MIP) models, are inefficient and inaccurate, leading to delays and incorrect resource allocation in computing systems, which affects performance and user experience.
Innovation Solution
A machine learning-based approach that generates distribution thresholds using classification models to segment entities based on the likelihood of a particular event, allowing for efficient and accurate allocation of resources by categorizing entities into segments with different intervention protocols.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If mixed integer programming (MIP) models are used for entity population segmentation, then segmentation accuracy may be maintained, but segmentation time becomes excessively long
Solution Approach 1:
The patent replaces the mechanical optimization system (MIP models) with a machine learning-based system that uses trained classifiers and distribution analysis to determine segmentation thresholds. This substitution transforms the segmentation process from a computationally intensive optimization problem into a faster statistical analysis task, resolving the contradiction between accuracy and time.
Solution Approach 2:
The patent performs preliminary actions by pre-training machine learning classifiers and pre-computing distribution characteristics using historical data before actual segmentation is needed. This allows the system to quickly segment new entity populations without performing time-consuming real-time optimization, while maintaining accuracy through the pre-learned patterns.
2Productivity
If existing population segmentation techniques are used, then some level of entity categorization is achieved, but resource allocation accuracy deteriorates
Solution Approach 1:
The patent incorporates feedback mechanisms by using labeled training data that reflects actual event occurrences and outcomes. The machine learning models are trained on this feedback to learn which segmentation thresholds lead to accurate predictions and effective resource allocation, continuously improving categorization precision based on observed results.
Solution Approach 2:
The patent dynamically adjusts segmentation parameters (thresholds) based on learned distributions and target event rates. Instead of using fixed or heuristic-based thresholds, the system optimizes threshold parameters to achieve both high categorization accuracy and effective resource allocation, resolving the contradiction between productivity and precision.
3Speed
If segmentation is performed quickly using simplified methods, then processing speed improves, but segmentation accuracy deteriorates
Solution Approach 1:
The patent replaces complex mechanical optimization processes with machine learning-based statistical analysis. The trained models can quickly evaluate new data points against learned distributions without performing iterative optimization, achieving both high speed and maintained accuracy through the substitution of the underlying computational mechanism.
Data Source
AI summary
Aspects of the present disclosure relate to machine learning-based techniques for segmenting entity populations. Embodiments include approximating distributions for outputs generated by a classification machine learning model and ground truth occurrences of a targeted event. A first distribution may relate to the distribution of classification model scores that indicate the likelihood of the targeted event occurring with respect to entities. A second distribution may relate to the distribution of actual occurrences of the targeted event. Based on the distributions, thresholds may be generated by minimizing the values of the thresholds as a function of the distributions and a targeted rate of occurrence for the targeted entity with respect to different segments that are included within the thresholds. Entities may then be segmented based on the thresholds, and interventions (such as resource allocations) may be applied based on the segmentation.