Comparison of Decision Tree and K-Nearest Neighbor Performance for Determining Mustahik

Abstract

The problem of poverty is one of the fundamental issues of concern to the Indonesian government. One of the methods used by Islam to alleviate poverty is through zakat from Badan Amil Zakat Nasional (BAZNAS). Currently, the distribution of zakat is divided into two, namely in the form of consumptive zakat and productive zakat. Productive zakat is aimed at people who need business capital. To assist zakat managers in managing their funds, a mechanism is needed that can process mustahik data so that it can be selected more quickly and precisely using data mining. In this research, the data mining methods that will be used are K-nearest neighbor (KNN) and Decision Tree. The dataset used in this research is data obtained from BAZNAS and has been preprocessed to obtain a dataset with 7 attributes and 144 records. Decision trees, KNN Manhattan, and KNN Euclidean are used to predict mustahik candidates who are worthy of receiving zakat. The performance of the third method was tested using AUC and confusion matrix namely Accuracy, Precision, Recall, and F1 in each dataset split scenario of 70%:30%, 75%:25%, and 80%:20%. Based on the number of false positive and false negative results, the best performance obtained is KNN Euclidean with a dataset division scenario of 80%:20%.