APPLICATION OF BACKPROPAGATION FOR FORECASTING OPEN UNEMPLOYMENT IN MAKASSAR CITY

Based on data from the Statistics Bureau of South Sulawesi Province, the open unemployment rate in Makassar City has remained consistently high over the past ten years, averaging 11.41%. This highlights a persistent labor market issue and positions Makassar as the leading contributor to the open unemployment rate in the province. To support effective policymaking and early intervention strategies, it is essential to forecast future unemployment trends based on historical data. Therefore, this study aims to forecast the open unemployment rate in Makassar City over the next five years using a machine learning approach. Among the available forecasting methods, the Backpropagation Artificial Neural Network (ANN) was selected due to its proven ability to model complex, non-linear relationships often found in socio-economic data. ANN is particularly effective in handling temporal dynamics without assuming linearity or stationarity, unlike traditional statistical models. In this study, the forecasting process involved data normalization, scenario-based data partitioning, ANN architecture design, and model training and testing. The model with the best performance consisted of 11 neurons in the input layer, 55 neurons in the hidden layer, and 1 neuron in the output layer, using 80% of the data for training and 20% for testing. This configuration yielded a forecasting accuracy of 91.896%, with a MAPE of 8.131% and an MSE of 0.003. The denormalized results forecast a steady decline in the open unemployment rate from 9.078% in 2023 to 7.248% in 2027, indicating a positive trend in employment. Nevertheless, it is important to acknowledge the limitations of forecasting models and the potential influence of external factors that may affect actual outcomes.