K-Means – Resilient Backpropagation Neural Network in Predicting Poverty Levels

In solving economic problems, the government has implemented several development policies. However, this policy is considered to be too centered on big cities. So, through this research it is hoped that it can provide an overview related to regional groups that fall into the poorer category so that the government can also provide accelerated development policies that are oriented towards improving the economy of residents in the area. This study aims to determine the results of classifying district/city poverty levels in Indonesia as a basis for classification for predictions and to classify district/city poverty levels based on influencing factors. The method used in this study is K-Means Clustering using the poverty depth index and poverty severity index variables, then proceed with using the Backpropagation Neural Network (BNN) algorithm using the GRDP, per capita expenditure, human development index, and mean years of schooling. The results obtained using the K-Means algorithm are that there are 42 districts/cities that belong to cluster 1 where this region has a poverty index depth and severity index value that is higher than the 472 districts/cities in cluster 2. Furthermore, the cluster results are used as response variables for classification with BNN. The accuracy of the model obtained is very high, which is equal to 98.06, so the model is very feasible to be used as a poverty rate prediction model based on the variables used.