Abstract:Cost-sensitive learning methods guaranteeing privacy are becoming crucial nowadays in many applications where increasing use of sensitive personal information is observed. However, there has no optimal learning scheme developed in the literature to learn cost-sensitive classifiers under constraint of enforcing differential privacy. Our approach is to first develop a unified framework for existing cost-sensitive learning methods by incorporating the weight constant and weight functions into the classical regularized empirical risk minimization framework. Then, we propose two privacy-preserving algorithms with output perturbation and objective perturbation methods, respectively, to be integrated with the cost-sensitive learning framework. We showcase how this general framework can be used analytically by deriving the privacy-preserving cost-sensitive extensions of logistic regression and support vector machine. Experimental evidence on both synthetic and real data sets verifies that the proposed algorithms can reduce the misclassification cost effectively while satisfying the privacy requirement. A theoretical investigation is also conducted, revealing a very interesting analytic relation, i.e., that the choice of the weight constant and weight functions does not only influence the Fisher-consistent property (population minimizer of expected risk with a specific loss function leads to the Bayes optimal decision rule) but also interacts with privacy-preserving levels to affect the performance of classifiers significantly.

Towards cost-sensitive learning for real-world applications

Learning with cost intervals.

On Multi-Class Cost-Sensitive Learning

Completely Determined Known at Test Phrase Unknown at Test Phrase , but Some Information is Known Completely Unknown Impreciseness of Costs Previous CSL

Cost Sensitive Support Vector Machines

The Influence of Class Imbalance on Cost-Sensitive Learning: an Empirical Study.

SVM-Based Cost-sensitive Classification Algorithm with Error Cost and Class-dependent Reject Cost

Large Margin Distribution Learning with Cost Interval and Unlabeled Data.

Privacy-Preserving Cost-Sensitive Learning

SVM-based Cost Sensitive Mining

RUE: A Robust Personalized Cost Assignment Strategy for Class Imbalance Cost-sensitive Learning

A Multiphase Cost-Sensitive Learning Method Based on the Multiclass Three-Way Decision-Theoretic Rough Set Model

Training cost-sensitive neural networks with methods addressing the class imbalance problem

Cost-Sensitive Deep Learning with Layer-Wise Cost Estimation

A Study on the Effiect of Class Distribution Using Cost-Sensitive Learning

Optimization of Selective Ensemble for Cost-Sensitive Classification: an Empirical Study

Adaptive learning cost‐sensitive convolutional neural network

Soft Methodology for Cost-and-error Sensitive Classification

Cost-Sensitive Large Margin Distribution Machine

Learning Only When It Matters: Cost-Aware Long-Tailed Classification

Cost-sensitive Regression Learning on Small Dataset Through Intra-Cluster Product Favoured Feature Selection