ML Ensemble for Entity Failure Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning approaches struggle to accurately predict an entity's ability to meet a target parameter, especially for new entities with limited historical data, as they rely on data from other entities and do not consider multiple predicted scenarios, leading to inaccurate predictions.
Innovation Solution
A machine learning model ensemble is developed, comprising multiple sub-models and a principal model that generates simulated adjustments to sub-values, which are then analyzed to compute a risk classifier's likelihood of the entity failing to meet the target parameter, providing a more accurate prediction by considering various scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing machine learning approaches use data from other entities to predict target parameter, then prediction can be made for new entities with limited historical data, but prediction accuracy deteriorates because multiple predicted scenarios are not considered
Solution Approach 1:
The system segments the prediction task into multiple independent ML sub-models, each predicting a different scenario outcome. Instead of using a single complex model, the approach divides the prediction into separate sub-models that each handle specific scenarios, then combines their outputs to form a comprehensive prediction with multiple simulated values.
Solution Approach 2:
The system adds a new dimension to the prediction by generating multiple simulated values for the target parameter instead of a single prediction. This dimensional expansion allows the system to capture uncertainty and provide a distribution of possible outcomes, significantly improving prediction accuracy for new entities.
2Reliability
If multiple ML sub-models are trained to generate simulated adjustments, then comprehensive scenario analysis is achieved, but computational complexity and training time increase
Solution Approach 1:
The system performs preliminary action by pre-training multiple ML sub-models on historical data before actual prediction is needed. This allows the sub-models to be ready and configured in advance, so when prediction is required, the system can quickly generate multiple simulated scenarios without extensive real-time computation.
Solution Approach 2:
The system uses parameter changes by adjusting input parameters to the sub-models to generate different simulated scenarios. By varying input parameters within reasonable ranges, the system creates multiple predicted outcomes without retraining the models, thus maintaining reliability while reducing computation time.
Data Source
AI summary
There is provided a method of generating a machine learning (ML) model ensemble for computing likelihood of an entity failing to meet a target parameter, comprising: training ML-sub-models that each output sub-values for an input of raw data elements, training a principal ML model that outputs a value of an entity parameter corresponding to the target parameter for an input of the sub-values, using a training dataset including for sample entities, the ML-sub-values and corresponding entity parameters, inputting raw data elements associated with the entity into the ML-sub-models to obtain respective sub-values, in iterations: computing simulated adjustments to the sub-values to generate adjusted sub-values that are inputted into the principal ML model to obtain simulated values for the entity parameter, and generating a risk classifier that generates a likelihood of the entity failing to meet the target parameter according to an analysis of the simulated values for the entity parameter.