Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Job Salary Prediction with Disentangled Composition Effect Modeling: A Neural Prototyping Approach

Yang Ji, Ying Sun, Hengshu Zhu

TL;DR

This work tackles salary prediction by modeling how skill compositions shape compensation, a problem where traditional methods struggle to capture set-level, multi-faceted semantics. It introduces LGDESetNet, an intrinsically explainable framework that jointly learns disentangled local subsets and globally influential prototypical skill sets, with optional set-wise context fusion. By combining a multi-view, discrete subset selector with a set-oriented prototypical learner and graph-based regularization, the approach yields accurate salary predictions while providing interpretable, context-aware explanations of which skill sets drive pay. Empirical results across four real-world datasets, complemented by ablations, visualizations, a user study, and a case study, demonstrate both superior predictive performance and richer explanations of salary-influencing patterns, supporting its practical value for recruiters, job seekers, and policymakers.

Abstract

In the era of the knowledge economy, understanding how job skills influence salary is crucial for promoting recruitment with competitive salary systems and aligned salary expectations. Despite efforts on salary prediction based on job positions and talent demographics, there still lacks methods to effectively discern the set-structured skills' intricate composition effect on job salary. While recent advances in neural networks have significantly improved accurate set-based quantitative modeling, their lack of explainability hinders obtaining insights into the skills' composition effects. Indeed, model explanation for set data is challenging due to the combinatorial nature, rich semantics, and unique format. To this end, in this paper, we propose a novel intrinsically explainable set-based neural prototyping approach, namely \textbf{LGDESetNet}, for explainable salary prediction that can reveal disentangled skill sets that impact salary from both local and global perspectives. Specifically, we propose a skill graph-enhanced disentangled discrete subset selection layer to identify multi-faceted influential input subsets with varied semantics. Furthermore, we propose a set-oriented prototype learning method to extract globally influential prototypical sets. The resulting output is transparently derived from the semantic interplay between these input subsets and global prototypes. Extensive experiments on four real-world datasets demonstrate that our method achieves superior performance than state-of-the-art baselines in salary prediction while providing explainable insights into salary-influencing patterns.

Enhancing Job Salary Prediction with Disentangled Composition Effect Modeling: A Neural Prototyping Approach

TL;DR

This work tackles salary prediction by modeling how skill compositions shape compensation, a problem where traditional methods struggle to capture set-level, multi-faceted semantics. It introduces LGDESetNet, an intrinsically explainable framework that jointly learns disentangled local subsets and globally influential prototypical skill sets, with optional set-wise context fusion. By combining a multi-view, discrete subset selector with a set-oriented prototypical learner and graph-based regularization, the approach yields accurate salary predictions while providing interpretable, context-aware explanations of which skill sets drive pay. Empirical results across four real-world datasets, complemented by ablations, visualizations, a user study, and a case study, demonstrate both superior predictive performance and richer explanations of salary-influencing patterns, supporting its practical value for recruiters, job seekers, and policymakers.

Abstract

In the era of the knowledge economy, understanding how job skills influence salary is crucial for promoting recruitment with competitive salary systems and aligned salary expectations. Despite efforts on salary prediction based on job positions and talent demographics, there still lacks methods to effectively discern the set-structured skills' intricate composition effect on job salary. While recent advances in neural networks have significantly improved accurate set-based quantitative modeling, their lack of explainability hinders obtaining insights into the skills' composition effects. Indeed, model explanation for set data is challenging due to the combinatorial nature, rich semantics, and unique format. To this end, in this paper, we propose a novel intrinsically explainable set-based neural prototyping approach, namely \textbf{LGDESetNet}, for explainable salary prediction that can reveal disentangled skill sets that impact salary from both local and global perspectives. Specifically, we propose a skill graph-enhanced disentangled discrete subset selection layer to identify multi-faceted influential input subsets with varied semantics. Furthermore, we propose a set-oriented prototype learning method to extract globally influential prototypical sets. The resulting output is transparently derived from the semantic interplay between these input subsets and global prototypes. Extensive experiments on four real-world datasets demonstrate that our method achieves superior performance than state-of-the-art baselines in salary prediction while providing explainable insights into salary-influencing patterns.

Paper Structure

This paper contains 38 sections, 5 theorems, 17 equations, 15 figures, 6 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Data-driven Salary Prediction
  4. Set Modeling
  5. Neural Network Explainability
  6. Problem Formulation
  7. Methodology
  8. Framework Overview
  9. Disentangled Discrete Subset Selection
  10. Multi-view Subset Selection Network
  11. Skill-wise External Information Fusion
  12. Co-occurrence Graph Density Regularization
  13. Prototypical Set Learning
  14. Set-oriented Prototypical Part Layer
  15. Set-wise External Information Fusion
  16. ...and 23 more sections

Key Result

lemma 1

(Permutation matrix and permutation function) $\forall X \in \mathbb{R}^{N \times N}$, for all permutation matrix $P$ of size $N$, there exists $\pi: \{1,2,...,N\} \rightarrow \{1,2,...,N\}$ is a permutation function, which satisfies:

Figures (15)

  • Figure 1: An illustrative example on skill-salary influence.
  • Figure 2: The architecture of LGDESetNet.
  • Figure 4: Training procedure of LGDESetNet.
  • Figure 5: Ablation studies across four datasets.
  • Figure 6: Parameter sensitivity of IT job postings. Left: Prototype number $M$. Right: View number $H$.
  • ...and 10 more figures

Theorems & Definitions (9)

  • lemma 1
  • lemma 2
  • proof
  • lemma 3
  • proof
  • Proposition 1
  • proof
  • Proposition 2
  • proof