Kajian Algoritma C4.5 dan K-NN Untuk Memprediksi Penduduk Miskin

Abstract

Poverty is a serious issue that affects people all over the world. The United Nations (UN) has recognised poverty as a top priority in its Sustainable Development Goals. Although there has been a decline in poverty rates in recent years, there are still many individuals who struggle to fulfil their basic needs. Therefore, more effective efforts are needed in identifying the poor so that aid programmes can be well-targeted. This research aims to compare two classification methods, namely C4.5 and K-Nearest Neighbors (KNN), in predicting poverty rates. The C4.5 method uses a decision tree to classify the data, while KNN uses the closest distance to perform the classification. The data used in this research is poverty data in Indonesia. This research methodology involves data pre-processing stages, including data cleaning, feature selection, data exploration, and data balancing. Next, model training and testing using the C4.5 and KNN algorithms were conducted. The model performance is evaluated using metrics such as accuracy, recall, precision, F1 measure, and Area Under Curve (AUC). This research is still at the model design stage, and the follow-up will be to continue the research until the algorithm evaluation results. Using the confusion matrix, the best algorithm that can detect the poor with high accuracy will be selected. The results of this research are expected to provide useful insights in developing effective assistance programmes for poverty alleviation in Indonesia.

Keywords : Poverty; C.45; K-Nearest Neighbours (KNN), Prediction