| dc.description.abstract | When applying regression analysis, it is often influenced by spatial and temporal aspects. This will have an impact on forming a good model. Utilizing a geographically weighted panel regression approach, namely geographically weighted panel regression (GWPR) is one way to handle this. GWPR is a development of regression analysis that considers spatial and temporal aspects. When using the GWPR method, a weighting function is needed to weight each research area. The weighting function is divided into two, namely fixed kernel and adaptive fixed kernel. The two weights are differentiated based on the bandwidth value at each location. Bandwidth is a fixed distance or radius from one observation point to another point. One of the many provinces in Indonesia that is trying to overcome the problem of open unemployment is North Sumatra. The Open Unemployment Rate (TPT) in North Sumatra Province was recorded at around 6.91% in 2020. This percentage is still greater than other provinces on the island of Sumatra, such as South Sumatra with 5.51% and Jambi 5.13%. This study uses secondary data taken from BPS, including cross section data from 33 districts/cities in North Sumatra and time series data from 2018 to 2022. The optimum weighting matrix is produced by the Gaussian kernel weighting function. In the model adjustment test process which was carried out simultaneously, the GPWR model had better goodness of fit when compared to the global regression model. The research results show the statistical value of the t test, standard error and p-value for each variable at each observation location through the use of partial testing. GWPR modeling produces an R2 value of 93,511% and an RMSE of 0,6702. Meanwhile, the results obtained with global modeling, FEM panel regression with within estimator gave an R2 value of 92,72% and an RMSE of 0,9060. Thus, it can be concluded that the use of the GWPR model is better than the global model. | en_US |
