Paper Type

Complete

Paper Number

PACIS2025-1845

Description

In the digital era, job seekers increasingly rely on online job reviews to make informed career decisions. However, the sheer volume of reviews can lead to information overload, making it difficult to identify the most helpful insights. This study proposes a machine learning-based approach to predict the helpfulness of job reviews on Glassdoor. By integrating reviewer employment status, firm-level variables, and thematic content extracted using LDA topic modeling, we construct and compare Random Forest and XGBoost models. The Random Forest model achieves superior accuracy, and SHAP analysis identifies key factors influencing helpfulness, such as review length (particularly in the cons section) and topics related to poor management, salary concerns, and job challenges. The findings provide practical implications for improving review ranking mechanisms and enhancing user experience on online job platforms.

Comments

AI ML

Download

Share

COinS
 
Jul 6th, 12:00 AM

Predicting Job Review Helpfulness: Machine Learning and SHAP Approach

In the digital era, job seekers increasingly rely on online job reviews to make informed career decisions. However, the sheer volume of reviews can lead to information overload, making it difficult to identify the most helpful insights. This study proposes a machine learning-based approach to predict the helpfulness of job reviews on Glassdoor. By integrating reviewer employment status, firm-level variables, and thematic content extracted using LDA topic modeling, we construct and compare Random Forest and XGBoost models. The Random Forest model achieves superior accuracy, and SHAP analysis identifies key factors influencing helpfulness, such as review length (particularly in the cons section) and topics related to poor management, salary concerns, and job challenges. The findings provide practical implications for improving review ranking mechanisms and enhancing user experience on online job platforms.