Machine Learning Model Training With Feasibility-Constrained Proposals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models struggle to provide improvement proposals that achieve both high prediction accuracy and feasibility, often resulting in unfeasible suggestions like making an age younger to obtain desired loan approval.
Innovation Solution
A training program that calculates an evaluation index based on prediction accuracy and feasibility of improvement proposals, using training data and constraint conditions to train machine learning models, allowing for feasible suggestions while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are trained to maximize prediction accuracy, then the model can correctly predict outcomes, but the improvement proposals generated become unfeasible (e.g., suggesting to make age younger)
Solution Approach 1:
The patent changes the training parameters of the machine learning model by introducing a new objective function that combines prediction accuracy with feasibility constraints. Instead of only maximizing prediction accuracy, the model is trained to optimize a composite objective that includes feasibility metrics, thereby generating proposals that are both accurate and practically implementable.
Solution Approach 2:
The patent introduces constraint conditions as intermediary elements that mediate between the prediction objective and the feasibility requirement. These constraints act as a bridge, guiding the model to generate proposals that satisfy both prediction accuracy and feasibility criteria by incorporating domain-specific knowledge into the training process.
2Productivity
If the machine learning model focuses on generating any improvement proposal, then it can provide suggestions, but the proposals lack consideration of realization possibilities
Solution Approach 1:
The patent implements feedback mechanisms during model training by incorporating feasibility assessments into the loss function. The model receives feedback on whether generated proposals are feasible based on constraint conditions, and adjusts its parameters accordingly. This feedback loop ensures that the model learns to generate not only numerous proposals but also reliable, realizable ones.
Solution Approach 2:
The patent applies preliminary action by pre-defining constraint conditions and feasibility criteria before the model generates proposals. These preliminary specifications guide the model's proposal generation process, ensuring that feasibility considerations are built into the generation mechanism from the outset rather than being evaluated afterward.
Data Source
AI summary
A non-transitory computer-readable recording medium stores therein a training program of a machine learning model that outputs a proposal for obtaining a desired result, the training program of a machine learning model causes a computer to execute a process including acquiring training data including a plurality of attributes, acquiring constraint condition data of the attributes, calculating first information regarding prediction accuracy of the machine learning model based on the training data, calculating second information regarding feasibility of the proposal based on the training data and the constraint condition data, calculating an evaluation index based on the first information and the second information, and training the machine learning model based on the evaluation index.