Ordinal Time Series Classification with Missing Labels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in accurately classifying ordinal time series data with missing labels, as they struggle to preserve the order of labels in the feature space and handle missing information effectively.
Innovation Solution
The proposed solution involves a computer-implemented method that uses a neural network model to compute rank-based statistics for incoming time series segments, selects candidate labels, and applies k-NN-based classification or hypothesis testing to classify the segments, even when missing labels are present.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional classification methods are used for ordinal time series data, then the classification process is simple, but the accuracy deteriorates when missing labels are present
Solution Approach 1:
The classification method is segmented into multiple conditional pathways based on the presence of candidate labels in training data. When both candidate labels are present, k-NN classification is applied; when one is missing, hypothesis testing is used; when both are missing, rank-based statistics are applied. This segmentation allows the system to handle different scenarios with appropriate methods, improving accuracy without applying unnecessary complexity to each case.
Solution Approach 2:
The system performs preliminary computation of rank-based statistics and selection of candidate labels before applying the final classification method. This preliminary action prepares the data and identifies the appropriate classification pathway in advance, enabling the system to handle missing labels effectively while maintaining a structured approach to the classification process.
2Measurement precision
If rank-based statistics and multiple classification methods are applied, then classification accuracy with missing labels is improved, but computational complexity increases
Solution Approach 1:
Different classification methods are applied locally based on the specific scenario encountered. The system computes rank-based statistics and candidate labels, then selectively applies k-NN classification, hypothesis testing, or rank-based classification depending on which candidate labels are present in the training data. This local quality approach ensures computational resources are used efficiently for each specific case rather than applying a uniformly complex method to all data.
3Reliability
If the system handles missing labels through multiple classification pathways, then robustness is improved, but the ease of operation deteriorates
Solution Approach 1:
The system introduces an intermediary classification module that acts as a mediator between the input time series data and the final classification output. This module automatically determines which classification pathway to follow based on the presence of candidate labels, handling the complexity of multiple methods internally while presenting a unified interface for operation. The intermediary manages the conditional logic and method selection, making the system robust to missing labels without requiring the user to manually manage the complexity of multiple classification pathways.
Data Source
AI summary
A method classifies missing labels. The method computes, using a neural network model trained on training data, rank-based statistics of a feature of a time series segment to attempt to select two candidate labels from the training data that the segment most likely belongs to. The method classifies the segment using k-NN-based classification applied to the training data, responsive to the two candidate labels being present in the training data. The method classifies the segment by hypothesis testing, responsive to only one candidate label being present in the training data. The method classifies the segment into a class with higher values of the rank-based statistics from among a plurality of classes with different values of the rank-based statistics, responsive to no candidate labels being present in the training data. The method corrects a prediction by an applicable one of the classifying steps by majority voting with time windows.