Semiconductor Yield Factor Ranking Across AI Prediction Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing semiconductor yield prediction models lack consistency in determining the contribution of predictive factors across different artificial intelligence models, making it difficult to identify key factors affecting yield in semiconductor manufacturing processes.
Innovation Solution
Implementing a SHAP (Shapley Additive exPlanations) algorithm to analyze the contribution of predictive factors in multiple AI models, scaling the contributions to a common range, and determining a contribution rank based on weighted averages or ranks to provide consistent and reliable insights into factor importance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If multiple artificial intelligence models are used to predict semiconductor yield, then the comprehensiveness of prediction factors is improved, but the consistency and reliability of contribution analysis deteriorates
Solution Approach 1:
The patent combines multiple AI models (random forest, gradient boosting, neural networks) into an integrated analysis framework that processes predictions from all models simultaneously. By merging the contribution analyses from different models and applying normalization and ranking techniques, the system achieves consistent identification of key predictive factors across diverse model architectures, thus resolving the contradiction between comprehensiveness and consistency.
Solution Approach 2:
The patent transforms the contribution values from different AI models through parameter changes including normalization to a common scale and ranking transformation. This parameter transformation enables direct comparison and aggregation of contribution metrics across models with different output ranges and scales, ensuring reliable and consistent factor importance ranking despite model diversity.
2Measurement precision
If prediction contributions from various AI models are aggregated, then the accuracy of factor identification is improved, but the complexity of the analysis system increases
Solution Approach 1:
The patent segments the complex analysis process into distinct modular steps: (1) individual model prediction and contribution calculation, (2) normalization of contribution values to unified scales, (3) aggregation of normalized contributions across models, and (4) ranking of factors by aggregated contribution. This segmentation reduces system complexity by making each step independent and manageable while maintaining overall accuracy.
Solution Approach 2:
The patent introduces normalization as an intermediary process between individual model predictions and final aggregation. This intermediary step transforms diverse contribution metrics into a common language, enabling accurate aggregation without directly combining heterogeneous model outputs, thus simplifying the overall analysis system.
Data Source
AI summary
A semiconductor yield prediction model analysis method according to one embodiment of the present disclosure may be performed by a computing device, and may comprise selecting a plurality of models having a prediction accuracy exceeding a preset threshold value, wherein each of the plurality of models is an artificial intelligence model configured to receive semiconductor-related data as a predictive factor and predict a semiconductor yield using the predictive factor; selecting a plurality of predictive factors commonly included in the plurality of models; obtaining a prediction contribution of each of the selected plurality of predictive factors to a prediction result of each of the plurality of models; scaling the prediction contributions of the predictive factors obtained on each of the models so as to be within the same range; and calculating a contribution rank of each of the selected plurality of predictive factors based on the scaled prediction contributions.