A comparative analysis of job satisfaction prediction models using machine learning: a mixed-method approach

Abstract

Purpose

Because the food service industry is more dependent on customer contact and human resources than other industries, it is crucial to understand the factors influencing employee job satisfaction to ensure that employees provide satisfactory service to customers. However, few studies have incorporated employee reviews of job portals into their research. Many job seekers tend to trust company reviews posted by employees on job portals based on the information provided by the company itself. Thus, this study utilized company reviews and job satisfaction ratings from employees in the food service industry on a job portal site, Job Planet, to conduct mixed-method research.

Design/methodology/approach

For qualitative research, we applied the Latent Dirichlet Allocation (LDA) model to food service industry company reviews to identify 10 job satisfaction factors considered important by employees. For quantitative research, four algorithms were used to predict job satisfaction ratings: regression tree, multilayer perceptron (MLP), random forest and XGBoost. Thus, we generated predictor variables for six cases using the probability values of topics and job satisfaction ratings on a five-point scale through LDA and used them to build prediction algorithms.

Findings

The analysis showed that algorithm accuracy performed differently in each of the six cases, and overall, factors such as work-life balance and work environment have a significant impact on predicting job satisfaction ratings.

Originality/value

This study is significant because its methodology and results suggest a new approach based on data analysis in the field of human resources, which can contribute to the operation and planning of corporate human resources management in the future.

Citation

Kim, J., Chang, P.-S., Yang, S.-B., Choi, I. and Lee, B. (2024), "A comparative analysis of job satisfaction prediction models using machine learning: a mixed-method approach", Data Technologies and Applications, Vol. ahead-of-print No. ahead-of-print. https://doi.org/10.1108/DTA-10-2023-0697