Can Large Language Model Predict Employee Attrition?

Xiaoye Ma, Weiheng Liu, Changyi Zhao, Liliya R. Tukhvatulina · 2024

Read Can Large Language Model Predict Employee Attrition? by Xiaoye Ma, Weiheng Liu, Changyi Zhao, Liliya R. Tukhvatulina (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2411-01353.txt.

Description

Abstract . Employee attrition is a critical issue faced by organizations, with significant costs associated with turnover and the loss of valuable talent. Traditional methods for predicting attrition often rely on statistical techniques that, while useful, struggle to capture the complexity of modern workforces.

Catalog record

CREATORXiaoye Ma, Weiheng Liu, Changyi Zhao, Liliya R. Tukhvatulina
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE37 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2411-01353

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Can Large Language Model Predict Employee Attrition? Xiaoye Ma, Weiheng Liu, Changyi Zhao, Liliya R. Tukhvatulina Abstract . Employee attrition is a critical issue faced by organizations, with significant costs associated with turnover and the loss of valuable talent. Traditional methods for predicting attrition often rely on statistical techniques that, while useful, struggle to capture the complexity of modern workforces. Recent advancements in machine learning (ML) have provided more accurate, scalable solutions, allowing organizations to analyze diverse data points and predict attrition with greater precision. However, the emergence of large language models (LLMs) has opened new possibilities in human resource management by offering the ability to interpret contextual information from employee communications and detect subtle cues related to turnover. In this paper, we leverage the IBM HR Analytics Employee Attrition dataset to evaluate the effectiveness of a fine-tuned GPT-3.5 model in comparison to traditional machine learning classifiers, including Logistic Regression, k-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree, Random Forest, AdaBoost, and XGBoost. Our study focuses on assessing the predictive power, interpretability, and real-world applicability of each model. While traditional models offer ease of use and transparency, LLMs have the potential to uncover more nuanced patterns in employee behavior. Through our analysis, we aim to provide practical insights for organizations seeking to enhance their employee retention strategies with advanced predictive tools. Our results show that the fine-tuned GPT-3.5 large language model (LLM) outperforms traditional machine learning approaches in terms of prediction accuracy, with an impressive precision of 0.91, recall of 0.94, and an F1-score of 0.92. In contrast, the best-performing traditional model, Support Vector Machine (SVM), achieved an F1-score of 0.82, while ensemble methods like Random Forest and XGBoost reached F1-scores of 0.80. These findings highlight the ability of GPT-3.5 to capture complex patterns in employee behavior and attrition risks, offering enhanced interpretability by identifying subtle linguistic cues and recurring themes. This demonstrates the potential of integrating LLMs into HR strategies to significantly improve predictive performance and decision-making in employee retention. DOI: XXXXXXX.XXXXXXX Conference: Make sure to enter the correct conference title from your rights confirmation emai; June 03–05, 2018; Woodstock, NY ISBN: 978-1-4503-XXXX-X/18/06 CCS: Computing methodologies CCS: Computing methodologies Machine learning 1. Introduction Employees are invaluable assets and investments for companies of all sizes, from small businesses to large multinational corporations. For instance, the average cost of a single interview in Switzerland is CHF 400, and the median cost to hire a skilled employee can reach CHF 9,738 ( Blatter et al., 2016 ) . While recruitment processes are designed to select the most suitable candidates, they do not always guarantee successful outcomes. Employee attrition has long been a challenge for human resource managers and employers, particularly in fast-growing economies ( Iqbal, 2010 ; Alao and Adeyemo, 2013 ) . Employee attrition is a multifaceted issue influenced by various factors that impact organizational performance. Common causes include the absence of career growth opportunities, inadequate compensation, and poor work-life balance, all of which contribute to dissatisfaction and burnout ( Latha, 2013 ; Alao and Adeyemo, 2013 ) . Additionally, mismanagement, a mismatch between job responsibilities and employee skills, and cultural disconnects within the organization can further exacerbate turnover ( Pallathadka et al., 2022 ) . Employees may also leave due to better job offers, lack of recognition, unresolved workplace conflicts, or perceived job insecurity ( Raza et al., 2022 ; Burgard et al., 2009 ) . Understanding these factors is crucial for organizations aiming to improve employee retention strategies. Traditional approaches to predicting employee attrition typically involve the analysis of historical data using statistical methods to identify patterns and trends. These methods often require manual data analysis, such as reviewing exit interviews, employee surveys, and demographic information to detect common reasons for leaving. HR teams tend to focus on key indicators such as job tenure, performance reviews, and salary growth to assess the risk of turnover. However, these methods can be limited by their reliance on retrospective data and may struggle to capture the complexities of attrition dynamics in larger, more diverse workforces. They may also face challenges related to scalability and accuracy. Despite these limitations, traditional methods have served as foundational tools for understanding workforce trends and informing retention strategies before the advent of more advanced machine learning techniques. To address these challenges, machine learning (ML) has gained traction as a more powerful method for employee attrition prediction. ML models can analyze vast amounts of data, uncover complex patterns, and make more accurate predictions than traditional approaches. By incorporating diverse data points, such as employee engagement levels, work performance, demographic characteristics, and external factors like market trends, ML models offer a more comprehensive view of attrition risks. These models—including classification algorithms like Logistic Regression, Decision Trees, Random Forests, and advanced techniques like XGBoost and neural networks—can continuously learn from new data, improving predictive performance over time. Moreover, ML-based systems can handle large-scale, real-time data, making them highly adaptable to dynamic workforce environments. These systems can also detect early warning signs of attrition, enabling organizations to intervene proactively. By leveraging these predictive capabilities, businesses can take targeted actions, such as personalizing employee development programs, adjusting compensation structures, or addressing disengagement issues. In this way, ML offers a forward-looking, data-driven solution that empowers organizations to mitigate employee turnover and maintain a competitive edge in talent retention. Recently, large language models (LLMs) have emerged as powerful tools in diverse fields, from natural language processing ( Achiam et al., 2023 ; Team et al., 2023 ; Touvron et al., 2023 ) to scientific research ( Chen et al., 2023b ; Ding et al., 2023 ; Chen et al., 2023a ) , and their applications are now expanding into human resource management. LLMs enable organizations to go beyond numerical data and tap into the contextual richness present in employee communications. These models can analyze subtle linguistic patterns, shifts in tone, or recurring themes in employee communications to detect potential attrition risks. This deeper understanding of employee sentiment allows HR professionals to make more informed decisions about retention strategies. As LLMs continue to advance, their integration into employee attrition prediction models offers the potential to significantly enhance accuracy, scalability, and interpretability, thereby transforming workforce retention practices. Despite the advantages of LLMs, their effectiveness in employee attrition prediction is still under scrutiny. In this work, we use the widely recognized IBM HR Analytics Employee Attrition dataset to compare the performance of a fine-tuned GPT-3.5 model with classic machine learning models, including Logistic Regression, k-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree, Random Forest, AdaBoost, and XGBoost. Our objective is to evaluate the predictive capabilities of the LLM against these traditional classifiers, with a particular focus on accuracy, interpretability, and real-world applicability. While traditional models have long been favored for attrition prediction due to their simplicity and interpretability, LLMs have the potential to uncover more complex patterns in employee behavior. This study aims to determine whether fine-tuned LLMs can outperform or complement traditional methods and provide unique insights into the drivers of employee turnover. By evaluating the strengths and weaknesses of each model, we aim to offer practical guidance for organizations seeking to integrate advanced predictive tools into their human resource strategies. Additionally, we explore

Read the complete text