Parameter estimation of multivariate adaptive regression spline (MARS) - Longitudinal (Case study: Open unemployment rate in East Java)

An important issue highlighted in the eighth goal of the Sustainable Development Goals (SDGs) is unemployment, which can be measured by the open unemployment rate (TPT) in a given region. From 2020 to 2023, the TPT in East Java Province has shown a declining trend. However, despite this decrease, the TPT in East Java over the past three years has not yet achieved the targets set in the Regional Government Work Plan (RKPD) for East Java Province. Unemployment is a highly complex, multidimensional (multivariate) problem. The ambiguous relationship between response and predictor variables makes traditional parametric statistical methods less suitable for analysis. In this context, the Multivariate Adaptive Regression Spline (MARS) method has emerged as a useful tool for handling high-dimensional data and has been extended to longitudinal (time-series) data analysis. Therefore, this study applies the MARS-Longitudinal method to identify factors influencing TPT from 2020 to 2023. The findings indicate that the three most significant variables affecting the open unemployment rate in East Java Province are the Human Development Index, Population Density, Gross Regional Domestic Product (GRDP) with the best R² value and RMSEA value shown by modeling using MARS-Longitudinal.