When LLM meets Fuzzy-TOPSIS for Personnel Selection through Automated Profile Analysis
By: Shahria Hoque, Ahmed Akib Jawad Karim, Md. Golam Rabiul Alam, Nirjhar Gope
Published: 2026-01-29
View on arXiv →Abstract
This study introduces an automated personnel selection system that combines large language models (LLMs) with Fuzzy-TOPSIS to enhance the hiring process. The system uses advanced natural language processing (NLP) to assess and rank software engineering applicants based on aggregated LinkedIn profiles, which include skills, education, and experience. Fuzzy-TOPSIS is then employed to handle the inherent uncertainties and vagueness in human evaluation, providing a robust and objective ranking. This approach aims to streamline recruitment, minimize human bias, and improve the efficiency of identifying top talent, offering a practical solution for contemporary human resource challenges.
Impact
practical
Topics
7
💡 Simple Explanation
Hiring people is hard because reading hundreds of resumes takes a long time, and recruiters might be biased. This paper proposes a computer system that uses 'Smart AI' (like ChatGPT) to read the resumes and understand the skills, and then uses a mathematical formula (Fuzzy-TOPSIS) to rank the candidates fairly. Instead of just guessing who is best, the system calculates exactly how close each person is to the 'perfect' candidate, helping companies hire faster and more fairly.
🎯 Problem Statement
Manual personnel selection is inefficient, prone to cognitive bias, and inconsistent. Traditional automated systems (ATS) often rely on rigid keyword matching, missing qualified candidates who use different terminology. Pure LLM-based evaluation lacks transparency and mathematical rigor in ranking.
🔬 Methodology
The authors propose a multi-stage framework. 1. Preprocessing: Resumes are converted to text. 2. Criteria Extraction: An LLM is prompted to analyze the text and assign linguistic ratings (e.g., 'Very Good', 'Poor') to predefined criteria like Education, Experience, and Skills. 3. Transformation: These linguistic ratings are converted into Triangular Fuzzy Numbers to handle ambiguity. 4. Ranking: The TOPSIS method computes the closeness coefficient of each candidate to the Ideal Positive Solution. The final output is a ranked list.
📊 Results
The proposed hybrid system demonstrated a higher correlation with human expert rankings compared to keyword-based ATS and standalone LLM scoring. It significantly reduced processing time (from hours to minutes for batches of CVs) and showed improved consistency, meaning the same CV received the same rank across multiple runs, which is often a challenge with stochastic LLMs.
✨ Key Takeaways
Combining the linguistic power of LLMs with the mathematical stability of Fuzzy-TOPSIS creates a powerful tool for structured decision-making from unstructured data. This approach solves the 'black box' problem of AI in HR by providing a traceable calculation path for every candidate's ranking.
🔍 Critical Analysis
The paper presents a solid 'Applied AI' use case. The combination of LLM for semantic processing and Fuzzy-TOPSIS for logical ranking is a smart architectural choice that mitigates the unpredictability of pure LLM agents. However, the paper might underestimate the cost implications of processing large volumes of documents through high-end LLMs and the legal risks associated with automated profiling under GDPR. It represents a practical step forward but is not a fundamental AI breakthrough.
💰 Practical Applications
- B2B SaaS subscription for HR departments charged per job posting.
- API usage model: Charge ATS platforms per CV processed.
- Consulting services for companies to set up custom 'Fuzzy Weighting' models for their specific corporate culture.