Using Neural Network Auto-Regression to Forecast the Palestinian Unemployment Rate

The Palestinian labor market faces prolonged and significant challenges includinghigher rates of unemployment, which has major economic and socialimplications. This study aims to forecast the behavior of the unemploymentrate in Palestine utilizing quarterly unemployment data over the period from2001Q1 to 2023Q2. The data is divided into training (in-sample) and testing(out-of-sample) datasets. The study applies a neural network auto-regressionmodel (NNAR) to provide future predictions of the unemployment rate forthe next ten quarters (2023Q3 – 2025Q4). The forecasting performance wasassessed on both in-sample and out-of-sample datasets. The findings suggestthat the optimal model to predict the Palestinian unemployment ratewas NNAR(1,1,10)[4]. The study compared this model with auto-regressiveintegrated moving average (ARIMA) and Holt-Winter's (HW) methods andthe results revealed that the estimated NNAR model outperformed thesemodels. The findings indicate that the unemployment rate is expected toremain high with values oscillating between 23.8 to 28.1%. Hence, this studysuggests that unemployment is a chronic economic problem with strong seasonality.The findings provide valuable insights for policy implications andstrategies to address the unemployment issue in Palestine.