<![CDATA[The poverty rate in South Sulawesi Province remains a complex socio-economic issue that is closely linked to fluctuating unemployment levels each year. The imbalance between economic growth and labor absorption has increased the risk of poverty in several regions. Therefore, this study aims to analyze the effect of the unemployment rate on poverty in South Sulawesi Province through the development of a hybrid model based on machine learning and numerical methods, combining Support Vector Regression (SVR), Adams–Bashforth–Moulton (ABM), and Milne–Simpson approaches. The SVR model was employed to predict the Open Unemployment Rate (OUR), with the best performance achieved using the Radial Basis Function (RBF) kernel and k-fold = 8, resulting in a MAPE of 6.69% and R² of 0.92, indicating high predictive accuracy. Furthermore, the relationship between unemployment and poverty was formulated in the form of an ordinary differential equation (ODE) as follows: dp/dt= -1-0.195277U(t)-0.00907P(t). Where U(t) represents the unemployment rate and P(t) represents the poverty rate over time. Numerical simulations demonstrated that both the ABM and Milne–Simpson methods were capable of reconstructing the dynamic behavior of poverty levels with high accuracy, producing MAPE values of 2.09% and 4.04%, respectively. The results indicate that the hybrid SVR–ABM–Milne–Simpson model is effective in generating stable poverty predictions that closely approximate actual data. Among the two numerical methods, ABM outperformed Milne–Simpson, yielding smaller prediction errors and better numerical stability in terms of convergence behavior. In summary, this hybrid modeling framework provides a robust analytical and computational approach for understanding and forecasting the socio-economic interplay between unemployment and poverty in South Sulawesi.]]>
Copyrights © 2026
