Single-Stacked Support Vector Regression-Bayesian-Based Models for Unemployment Trend Analysis in Malaysia's Youth Labor Market

Abstract

In response to the evolving dynamics of the labor market shaped by natural disasters such as the Coronavirus Disease 2019 (COVID-19) pandemic, rapid technological progress, and the emergence of the Fourth Industrial Revolution (IR4.0), and given the very limited sample size of the unemployment dataset (less than 30 observations), this article introduces two advanced models that combine Support Vector Regression (SVR) with Bayesian statistical methods, namely the Single-Stacked SVR-Empirical Bayesian Beta-Binomial (SSVREBB) and the Single-Stacked SVR-Hierarchical Bayesian Binomial Proportion (SSVRHBB). These models are designed to handle data-limited environments while maintaining reliable predictive performance. They demonstrate superior flexibility in capturing both linear and nonlinear patterns and provide explicit quantification of prediction uncertainty, outperforming conventional time series, machine learning, and deep learning benchmarks. A VlseKriterijumska Optimizacija I Kompromisno Resenje (VIKOR)-based Taguchi optimisation framework is employed to identify the best-performing model configuration. The empirical analysis utilises annual unemployment data from 1995 to 2020, disaggregated by states, regions, and gender in Malaysia. The results indicate persistently high youth unemployment, particularly among females in Sabah (SBH), mainly due to gaps in education quality and skills mismatches. While the study focuses on the national context, it offers insights for policymakers aiming to reduce gender disparities, enhance female labor force participation, and support equitable employment opportunities in data-limited environments. In alignment with relevant United Nations Sustainable Development Goals (SDGs), specifically SDG1 (No Poverty), SDG4 (Quality Education), SDG5 (Gender Equality), and SDG8 (Decent Work and Economic Growth), this research demonstrates the potential of innovative predictive modelling to inform evidence-based labor market policies.