Chromatograph Mass Spectrometry Training Data Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for training models using peak information from chromatographs face accuracy and efficiency issues due to missing values and multicollinearity in peak data, which affect the performance of machine learning algorithms.
Innovation Solution
A training data generation apparatus that preprocesses peak information by deleting missing values, reducing correlated peaks, and removing outliers to generate high-quality training data for machine learning models, improving model accuracy and training efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If all peak information from multiple samples is used for training, then the quantity of training data is increased, but the accuracy of the trained model deteriorates due to missing values and multicollinearity
Solution Approach 1:
The patent extracts and removes problematic data elements (missing values and highly correlated peaks) from the training dataset. By identifying and eliminating peaks with correlation coefficients above a threshold value and handling missing values through imputation or removal, the system extracts only the useful, independent features while discarding redundant or erroneous data that degrades model accuracy.
Solution Approach 2:
The patent transforms the training data by applying parameter changes to handle missing values (e.g., imputation methods) and by selecting peaks based on correlation threshold parameters. This transformation converts raw peak information into processed training data with improved quality, where redundant correlated features are removed and missing data is appropriately handled, thereby improving model accuracy.
2Loss of information
If redundant correlated peaks are included in training data, then the completeness of information is maintained, but the training efficiency deteriorates due to multicollinearity
Solution Approach 1:
The patent extracts and removes redundant correlated peaks from the training dataset by calculating correlation coefficients between peaks and eliminating those above a predetermined threshold. This extraction process removes redundant information while maintaining the essential unique features, thereby improving training efficiency without significant loss of informative content.
Solution Approach 2:
The patent discards redundant correlated peaks that provide duplicate information, and recovers or maintains the essential unique features through careful selection. By discarding highly correlated peaks and keeping representative ones, the system optimizes the training dataset to improve efficiency while preserving the necessary information for accurate classification.
Data Source
AI summary
A processor performs processing for generating training data by processing a plurality of pieces of peak information obtained by a data obtaining unit. The processor deletes data on a peak missing in any of the plurality of pieces of peak information from each piece of peak information. When a coefficient of correlation of data between peaks among remaining peaks is equal to or larger than a prescribed value, the processor further deletes data on one peak of the peaks from each piece of peak information, and defines peak information including data on the remaining peaks as input data for data for learning.