Distributed Machine Learning for Database Element Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Health plan providers face challenges in efficiently enrolling individuals and employers in health insurance plans and providing targeted preventive healthcare information, while machine learning models struggle to accurately predict outcomes from large datasets and identify relevant relationships among variables.
Innovation Solution
A computerized method using machine learning models is employed to process historical feature vector inputs, generate service, selection, segment, title, background, decision, and retirement score outputs, and distribute campaign data based on these scores, utilizing structured databases and supervised learning techniques to classify entities and identify targets for targeted communication.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to predict outputs from large input datasets, then prediction accuracy and relationship identification improve, but processing time and computational complexity increase
Solution Approach 1:
The patent segments the machine learning processing into distributed components across multiple processing nodes. Historical feature vector inputs are processed in parallel by multiple nodes, each contributing to the overall prediction output. This segmentation allows the system to handle large datasets more efficiently by dividing the computational workload rather than processing everything sequentially on a single node.
Solution Approach 2:
The patent introduces distributed processing nodes as intermediaries between the input data and final prediction outputs. These intermediary processing nodes facilitate the transformation and analysis of historical feature vectors, enabling complex predictions to be made through coordinated computation across multiple nodes rather than requiring excessive computational power from a single source.
2Measurement precision
If machine learning models process historical feature vector inputs to generate multiple score outputs, then identification precision improves, but device complexity increases
Solution Approach 1:
The patent implements a universal machine learning model framework that can generate multiple different score outputs (service score, selection score, segment score, title score, background score, decision score, retirement score) from the same historical feature vector inputs. This multi-functional approach allows a single model architecture to serve multiple prediction purposes, reducing the need for separate specialized models for each scoring function.
Solution Approach 2:
The patent changes the output parameters of the machine learning model based on the specific prediction task at hand. By adjusting the model's output configuration, the same processing infrastructure can generate different types of scores (service likelihood, selection likelihood, retirement timing, etc.) without requiring fundamental changes to the underlying processing system or architecture.
3Productivity
If entities are classified and targeted for campaign distribution, then marketing effectiveness improves, but data processing requirements increase
Solution Approach 1:
The patent applies local quality by classifying entities into specific segments (service-providing, service-selecting, retirement-ready) and tailoring campaign distributions to each segment's specific characteristics. Rather than applying uniform marketing approaches to all entities, the system identifies and targets each segment with appropriate localized strategies, improving effectiveness while managing data processing by focusing resources on high-value segments.
Solution Approach 2:
The patent performs preliminary classification and scoring of entities before campaign distribution. By pre-processing the data to identify service-providing entities, likely service selectors, and retirement candidates in advance, the system prepares targeted audiences for subsequent campaign deployment. This preliminary action filters and organizes the data, reducing the processing burden during actual campaign execution while maintaining high targeting accuracy.
Data Source
AI summary
A computerized method of automatic distributed communication includes training a first and second machine learning models with historical feature vector inputs to generate a likelihood output and a mean count output, respectively. For each entity in a set, the method includes processing a likelihood feature vector input with the first machine learning model to generate a likelihood output indicative of a likelihood that the entity will have an avoidable negative health event within a specified first time period, and processing a mean count feature vector input with the second machine learning model to generate a mean count output indicative of an expected number of avoidable negative health events that the entity will have within a specified second time period. The method includes automatically distributing structured campaign data to at least a subset of entities in the set according to the likelihood output or the mean count output.