A Transfer-Based Additive LS-SVM Classifier for Handling Missing Data.

Area of Science:

• Machine Learning
• Data Science
• Computer Science

Background:

• Missing data significantly degrades classifier performance.
• Existing methods for handling missing data have limitations.
• Transfer learning offers a promising avenue for improving classification with incomplete data.

Purpose of the Study:

• To propose a novel transfer learning method for addressing missing data in classification.
• To enhance classification performance on incomplete training datasets.
• To introduce a method for identifying and mitigating classification errors from incomplete samples.

Main Methods:

• A novel transfer-based additive Least Squares Support Vector Machine (LS-SVM) classifier is proposed.
• Fast leave-one-out cross-validation is employed to assess sample influence and clean training data.
• The method is evaluated on seven public datasets and a community healthcare dataset.

Main Results:

• The proposed transfer-based additive LS-SVM classifier demonstrates comparable or superior performance to existing methods.
• It outperforms case deletion, mean imputation, and k-nearest neighbor imputation.
• The method effectively improves data quality by identifying influential incomplete samples.

Conclusions:

• Transfer learning provides an effective strategy for handling missing data in classification.
• The proposed transfer-based additive LS-SVM classifier is a robust solution for incomplete datasets.
• The method offers practical benefits for real-world applications, including healthcare data analysis.