Prediction Function Creation Device for Missing Value Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing prediction systems face reduced accuracy when dealing with a large number of variations in missing values for explanatory variables, as the increased number of labels results in smaller sample sizes for training, leading to lower prediction function accuracy.
Innovation Solution
A prediction function creation device and method that clusters samples based on patterns of missing values, assigns weights to these clusters, and uses partial training data to determine relevant explanatory variables for creating prediction functions tailored to each pattern of missing values, thereby increasing the sample size for each prediction function and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning methods are applied to training data with missing explanatory variable values, then prediction accuracy can be improved, but the methods become inapplicable when missing values are present
Solution Approach 1:
The patent converts the harmful effect of missing values (which make machine learning inapplicable) into a beneficial classification feature. By clustering samples based on their missing value patterns and treating these patterns as meaningful categories, the system enables machine learning to process and utilize data that would otherwise be unusable, thereby improving both applicability and prediction accuracy
2Adaptability or versatility
If missing values are imputed by mean value or predicted from other variables, then the data can be processed by machine learning, but unnecessary errors are introduced in the prediction
Solution Approach 1:
The patent segments the training data into distinct groups based on missing value patterns, creating separate subsets for each pattern type. This segmentation allows the system to apply appropriate processing to each group without introducing erroneous imputations, as each segment is handled according to its specific characteristics rather than applying a uniform imputation method that would propagate errors
3Adaptability or versatility
If data is partitioned by missing manner with each unique pattern receiving a separate label, then machine learning can be applied to each partition, but the sample size for each partition becomes small reducing prediction accuracy
Solution Approach 1:
The patent merges samples with similar missing value patterns into the same cluster, creating larger aggregated groups rather than treating each unique pattern separately. This merging increases the sample size for each cluster, enabling machine learning to achieve sufficient statistical power and prediction accuracy while still maintaining the ability to handle different missing value patterns through the clustering structure
Data Source
AI summary
The prediction function creation device according to the present invention for creating a prediction function to derive an objective variable by using a set of samples that include explanatory variables and an objective variable, the device includes: a clustering unit that clusters the respective samples by giving labels, and assigns weights to each label in accordance with patterns of missing values for the explanatory variables in labeled samples; a child model creation unit that makes portions of the training data partial training data on the basis of the weights, and determines an explanatory variable that constitutes the prediction function on the basis of patterns of missing values for the explanatory variables in the samples; and a mixture model creation unit that creates the prediction function with respect to each pattern of missing values by using the explanatory variable and the determined partial training data.