Predictive Prioritization via Supervised-Unsupervised Feature Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing predictive data analysis systems are inefficient and unreliable for performing predictive prioritization, particularly in scenarios with limited labeled training data, such as genetic test prioritization, where they struggle to accurately identify patients who would benefit from genetic testing and determine the appropriate tests.
Innovation Solution
The use of a combination of supervised and unsupervised machine learning models to generate target features for a resultant prioritization model, which eliminates the need for costly machine learning operations post-training, allowing for efficient and reliable predictive prioritization even with limited data. This involves identifying feature data objects, generating refined multi-input-type supervised machine learning models, and creating multi-dimensional unsupervised machine learning spaces to produce predicted prioritization scores.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If existing predictive data analysis systems are used for predictive prioritization, then the systems can perform basic predictions, but they are inefficient and unreliable particularly with limited labeled training data
Solution Approach 1:
The system segments the machine learning process into two distinct phases: an offline training phase that generates prioritization models and target features, and an online inference phase that efficiently scores new entities. This segmentation allows computationally intensive model training to be performed once, while deployment requires only lightweight scoring operations, thereby improving both reliability through proper model training and productivity through efficient online performance.
Solution Approach 2:
The system performs preliminary actions by pre-training machine learning models and pre-computing target features during an offline phase before actual predictive prioritization is needed. This preliminary model training and feature engineering enables the system to handle limited labeled data effectively while ensuring fast and reliable predictions during online operation without requiring repeated computationally intensive operations.
2Productivity
If supervised machine learning models are trained with limited labeled data, then the models can be trained faster, but the predictive accuracy and reliability decrease
Solution Approach 1:
The system introduces target features as an intermediary between the limited labeled training data and the final predictive prioritization scores. These target features are pre-computed during offline training and serve as enhanced inputs that capture complex relationships in the data, allowing the model to achieve high predictive accuracy even with limited labeled examples while maintaining fast training speed.
3Adaptability or versatility
If complex machine learning operations are performed post-training, then the models can adapt to new data, but the computational complexity and time consumption increase
Solution Approach 1:
The system performs all necessary model adaptation and feature engineering in advance during the offline training phase. The pre-trained models and pre-computed target features are stored and reused during online inference, eliminating the need for complex post-training operations during deployment. This approach maintains model adaptability through proper offline training while minimizing computational complexity during actual use.
Data Source
AI summary
Various embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and/or the like for performing predictive prioritization. Certain embodiments utilize systems, methods, and computer program products that perform predictive prioritization using a combination of supervised machine learning models and unsupervised machine learning models that are in turn used to generate target features for a resultant prioritization machine learning model.