Decision Making for Hierarchical Multi-label Classification with Multidimensional Local Precision Rate

  • 2022-05-16 18:43:35
  • Yuting Ye, Christine Ho, Ci-Ren Jiang, Wayne Tai Lee, Haiyan Huang
  • 0

Abstract

Hierarchical multi-label classification (HMC) has drawn increasing attentionin the past few decades. It is applicable when hierarchical relationships amongclasses are available and need to be incorporated along with the multi-labelclassification whereby each object is assigned to one or more classes. Thereare two key challenges in HMC: i) optimizing the classification accuracy, andmeanwhile ii) ensuring the given class hierarchy. To address these challenges,in this article, we introduce a new statistic called the multidimensional localprecision rate (mLPR) for each object in each class. We show thatclassification decisions made by simply sorting objects across classes indescending order of their true mLPRs can, in theory, ensure the class hierarchyand lead to the maximization of CATCH, an objective function we introduce thatis related to the area under a hit curve. This approach is the first of itskind that handles both challenges in one objective function without additionalconstraints, thanks to the desirable statistical properties of CATCH and mLPR.In practice, however, true mLPRs are not available. In response, we introduceHierRank, a new algorithm that maximizes an empirical version of CATCH usingestimated mLPRs while respecting the hierarchy. The performance of thisapproach was evaluated on a synthetic data set and two real data sets; ours wasfound to be superior to several comparison methods on evaluation criteria basedon metrics such as precision, recall, and $F_1$ score.

 

Quick Read (beta)

loading the full paper ...